OGMomentUnfolding.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "59c3fee9",
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import energyflow as ef\n",
    "import energyflow.archs\n",
    "from energyflow.archs import PFN\n",
    "from matplotlib import gridspec\n",
    "\n",
    "from matplotlib import rc\n",
    "rc('font', size=20)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "65d73cb5",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "<__array_function__ internals>:5: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray\n"
     ]
    }
   ],
   "source": [
    "#These are the same datasets from the OmniFold paper https://arxiv.org/abs/1911.09107.  More detail at https://energyflow.network/docs/datasets/.\n",
    "#Pythia and Herwig are two generators; one will be treated here as the \"simulation\" and one as \"data\".\n",
    "datasets = {'Pythia26': ef.zjets_delphes.load('Pythia26', num_data=1000000),\n",
    "            'Herwig': ef.zjets_delphes.load('Herwig', num_data=1000000)}"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "fb5376ea",
   "metadata": {},
   "source": [
    "## Set up the observables"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "609c1369",
   "metadata": {},
   "outputs": [],
   "source": [
    "def is_charged(myin):\n",
    "    if (myin == 0):\n",
    "        return 0\n",
    "    elif (myin == 0.1):\n",
    "        return 1\n",
    "    elif (myin == 0.2):\n",
    "        return -1\n",
    "    elif (myin == 0.3):\n",
    "        return 0\n",
    "    elif (myin == 0.4):\n",
    "        return -1\n",
    "    elif (myin == 0.5):\n",
    "        return 1\n",
    "    elif (myin == 0.6):\n",
    "        return -1\n",
    "    elif (myin == 0.7):\n",
    "        return 1\n",
    "    elif (myin == 0.8):\n",
    "        return 1\n",
    "    elif (myin == 0.9):\n",
    "        return -1\n",
    "    elif (myin == 1.0):\n",
    "        return 1\n",
    "    elif (myin == 1.1):\n",
    "        return -1\n",
    "    elif (myin == 1.2):\n",
    "        return 0\n",
    "    elif (myin == 1.3):\n",
    "        return 0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "a7c0258f",
   "metadata": {},
   "outputs": [],
   "source": [
    "for dataset in datasets:\n",
    "    mycharges = []\n",
    "    mycharges2 = []\n",
    "    for i in range(len(datasets[dataset]['gen_particles'])):\n",
    "        pTs = datasets[dataset]['gen_particles'][i][:,0]\n",
    "        charges = [is_charged(datasets[dataset]['gen_particles'][i][:,3][j]) for j in range(len(datasets[dataset]['gen_particles'][i][:,3]))]\n",
    "        mycharges+=[np.sum(charges*pTs**0.5)/np.sum(pTs**0.5)]\n",
    "        mycharges2+=[np.sum(np.abs(charges)*pTs)/np.sum(pTs)]\n",
    "    datasets[dataset]['gen_charge'] = mycharges\n",
    "    datasets[dataset]['gen_pTcharge'] = mycharges2\n",
    "\n",
    "    mycharges = []\n",
    "    mycharges2 = []\n",
    "    for i in range(len(datasets[dataset]['sim_particles'])):\n",
    "        pTs = datasets[dataset]['sim_particles'][i][:,0]\n",
    "        charges = [is_charged(datasets[dataset]['sim_particles'][i][:,3][j]) for j in range(len(datasets[dataset]['sim_particles'][i][:,3]))]\n",
    "        mycharges+=[np.sum(charges*pTs**0.5)/np.sum(pTs**0.5)]\n",
    "        mycharges2+=[np.sum(np.abs(charges)*pTs)/np.sum(pTs)]\n",
    "    datasets[dataset]['sim_charge'] = mycharges\n",
    "    datasets[dataset]['sim_pTcharge'] = mycharges2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "bec756e6",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "dict_keys(['gen_mults', 'gen_sdms', 'gen_tau2s', 'gen_zgs', 'gen_Zs', 'gen_ang2s', 'gen_lhas', 'gen_widths', 'gen_particles', 'gen_jets', 'sim_mults', 'sim_sdms', 'sim_tau2s', 'sim_zgs', 'sim_ang2s', 'sim_lhas', 'sim_widths', 'sim_particles', 'sim_jets', 'gen_charge', 'gen_pTcharge', 'sim_charge', 'sim_pTcharge'])"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "datasets['Pythia26'].keys()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "78eb3f21",
   "metadata": {},
   "outputs": [],
   "source": [
    "tau2s_Pythia_sim = datasets['Pythia26']['sim_tau2s']\n",
    "tau2s_Herwig_sim = datasets['Herwig']['sim_tau2s']\n",
    "\n",
    "tau1s_Pythia_sim = datasets['Pythia26']['sim_widths']\n",
    "tau1s_Herwig_sim = datasets['Herwig']['sim_widths']\n",
    "\n",
    "tau2s_Pythia_gen = datasets['Pythia26']['gen_tau2s']\n",
    "tau2s_Herwig_gen = datasets['Herwig']['gen_tau2s']\n",
    "\n",
    "tau1s_Pythia_gen = datasets['Pythia26']['gen_widths']\n",
    "tau1s_Herwig_gen = datasets['Herwig']['gen_widths']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "074eb4b8",
   "metadata": {},
   "outputs": [],
   "source": [
    "pT_true = datasets['Pythia26']['gen_jets'][:,0]\n",
    "m_true = datasets['Pythia26']['gen_jets'][:,3]\n",
    "pT_reco = datasets['Pythia26']['sim_jets'][:,0]\n",
    "m_reco = datasets['Pythia26']['sim_jets'][:,3]\n",
    "\n",
    "pT_true_alt = datasets['Herwig']['gen_jets'][:,0]\n",
    "m_true_alt = datasets['Herwig']['gen_jets'][:,3]\n",
    "pT_reco_alt = datasets['Herwig']['sim_jets'][:,0]\n",
    "m_reco_alt = datasets['Herwig']['sim_jets'][:,3]\n",
    "\n",
    "#\n",
    "w_true = datasets['Pythia26']['gen_widths']\n",
    "w_reco = datasets['Pythia26']['sim_widths']\n",
    "w_true_alt = datasets['Herwig']['gen_widths']\n",
    "w_reco_alt = datasets['Herwig']['sim_widths']\n",
    "\n",
    "#\n",
    "q_true = np.array(datasets['Pythia26']['gen_charge'])\n",
    "q_reco = np.array(datasets['Pythia26']['sim_charge'])\n",
    "q_true_alt = np.array(datasets['Herwig']['gen_charge'])\n",
    "q_reco_alt = np.array(datasets['Herwig']['sim_charge'])\n",
    "\n",
    "#\n",
    "r_true = np.array(datasets['Pythia26']['gen_pTcharge'])\n",
    "r_reco = np.array(datasets['Pythia26']['sim_pTcharge'])\n",
    "r_true_alt = np.array(datasets['Herwig']['gen_pTcharge'])\n",
    "r_reco_alt = np.array(datasets['Herwig']['sim_pTcharge'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "20740356",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.lines.Line2D at 0x7ff0c1515d00>"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.hist(r_true,bins=np.linspace(0,1.5,20))\n",
    "plt.axvline(0.66,color=\"black\",ls=\":\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8fdc8aa5",
   "metadata": {},
   "source": [
    "## Set up the binning \n",
    "\n",
    "Goal is to learn average value of the four jet substructure observables as a function of jet pT."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "ff438708",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "134 0.7082022305322636\n",
      "175 0.7078793549875664\n",
      "224 0.7075873586035227\n",
      "285 0.7093473767703654\n",
      "361 0.7091339517997708\n",
      "458 0.7075367685298829\n",
      "583 0.7074240849786398\n"
     ]
    }
   ],
   "source": [
    "#For the binning, make it so we have 50% in each bin growing from the low side\n",
    "binvals = [100]\n",
    "i = 0\n",
    "while binvals[-1] < 500:\n",
    "    for binhigh in range(binvals[i]+1,1000):\n",
    "        purity = len(pT_true[(pT_true > binvals[i])*(pT_true < binhigh)*(pT_reco > binvals[i])*(pT_reco < binhigh)]) / len(pT_true[(pT_true > binvals[i])*(pT_true < binhigh)])\n",
    "        if (purity > 0.5**0.5):\n",
    "            print(binhigh,purity)\n",
    "            i+=1\n",
    "            binvals+=[binhigh]\n",
    "            break"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "0f1cc93c",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig = plt.figure(figsize=(8, 6)) \n",
    "gs = gridspec.GridSpec(1, 1, height_ratios=[1]) \n",
    "ax0 = plt.subplot(gs[0])\n",
    "ax0.yaxis.set_ticks_position('both')\n",
    "ax0.xaxis.set_ticks_position('both')\n",
    "ax0.tick_params(direction=\"in\",which=\"both\")\n",
    "ax0.minorticks_on()\n",
    "plt.xticks(fontsize=20)\n",
    "plt.yticks(fontsize=20)\n",
    "\n",
    "n,b,_=plt.hist(pT_true,bins=binvals)\n",
    "plt.xlabel(\"Truth jet $p_T$ [GeV]\")\n",
    "plt.ylabel(\"Number of jets\")\n",
    "fig.savefig('figures/jetpt.pdf',bbox_inches='tight')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c5c72518",
   "metadata": {},
   "source": [
    "For the binned comparison, we pick the largest bins that have 50% purity."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "a711238b",
   "metadata": {},
   "outputs": [],
   "source": [
    "#Next, let's construct the response matrix.\n",
    "\n",
    "pTbin_truth = np.clip(np.digitize(pT_true,binvals),1,len(binvals)-1)-1\n",
    "pTbin_reco = np.clip(np.digitize(pT_reco,binvals),1,len(binvals)-1)-1\n",
    "\n",
    "#alt\n",
    "\n",
    "pTbin_truth_alt = np.clip(np.digitize(pT_true_alt,binvals),1,len(binvals)-1)-1\n",
    "pTbin_reco_alt = np.clip(np.digitize(pT_reco_alt,binvals),1,len(binvals)-1)-1"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "8e27fea0",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x432 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig = plt.figure(figsize=(8, 6)) \n",
    "gs = gridspec.GridSpec(1, 1, height_ratios=[1]) \n",
    "ax0 = plt.subplot(gs[0])\n",
    "ax0.yaxis.set_ticks_position('both')\n",
    "ax0.xaxis.set_ticks_position('both')\n",
    "ax0.tick_params(direction=\"in\",which=\"both\")\n",
    "ax0.minorticks_on()\n",
    "plt.xticks(fontsize=20)\n",
    "plt.yticks(fontsize=20)\n",
    "    \n",
    "H_pT, xedges, yedges = np.histogram2d(pTbin_truth,pTbin_reco,bins=[range(len(binvals)),range(len(binvals))])\n",
    "H_norm_pT = H_pT / H_pT.sum(axis=1, keepdims=True)\n",
    "plt.imshow(H_norm_pT,origin='lower',cmap=\"Reds\",vmin = 0,vmax = 1)\n",
    "cbar = plt.colorbar()\n",
    "cbar.ax.set_ylabel('Pr(Detector | Particle)') \n",
    "plt.xlabel(\"Particle-level $p_{T}$ bin\",fontsize=20)\n",
    "plt.ylabel(\"Detector-level $p_{T}$ bin\",fontsize=20)\n",
    "\n",
    "for i in range(len(binvals)-1):\n",
    "    for j2 in range(len(binvals)-1):\n",
    "        plt.text(j2,i, \"%0.2f\" % H_norm_pT.T[i,j2], \n",
    "                color=\"w\", ha=\"center\", va=\"center\", fontweight=\"bold\",fontsize=12)\n",
    "fig.savefig('figures/RsponsepT.pdf',bbox_inches='tight')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4e47414e",
   "metadata": {},
   "source": [
    "## Unfold"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "a9512935",
   "metadata": {},
   "outputs": [],
   "source": [
    "def IBU(T,D,R,n):\n",
    "    phis = [T]\n",
    "    for i in range(n):\n",
    "        m = R * phis[-1]\n",
    "        m /= (m.sum(axis=1)[:,np.newaxis] + 10**-50)\n",
    "        phis.append(np.dot(m.T, D))\n",
    "    return phis[-1]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "0e5dc37f",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([ 71857., 156028., 317371., 262234., 119543.,  48071.,  24896.])"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "T = np.sum(H_pT,axis=1)\n",
    "D = np.sum(H_pT,axis=0)\n",
    "IBU(T,D,H_norm_pT.T,1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "caefedd1",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([ 71857., 156028., 317371., 262234., 119543.,  48071.,  24896.])"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "T"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5e2acb08",
   "metadata": {},
   "source": [
    "## OmniFold\n",
    "\n",
    "(can skip past this part if you are only testing the Moment Unfolding parts, as this section takes some time to run)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "8eac0827",
   "metadata": {},
   "outputs": [],
   "source": [
    "from unfold import *\n",
    "\n",
    "import os\n",
    "os.environ['CUDA_VISIBLE_DEVICES']=\"1\"\n",
    "\n",
    "import tensorflow as tf\n",
    "import tensorflow.keras.backend as K\n",
    "tf.random.set_seed(1234)\n",
    "np.random.seed(1234)\n",
    "\n",
    "physical_devices = tf.config.list_physical_devices('GPU') \n",
    "tf.config.experimental.set_memory_growth(physical_devices[0], True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "02a629c5",
   "metadata": {},
   "outputs": [],
   "source": [
    "iterations = 5\n",
    "dims = 2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "34d30fbb",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.3313303940130997\n",
      "0.8936852369665991\n",
      "0.06441241362171127\n",
      "1.0481676174638868\n"
     ]
    }
   ],
   "source": [
    "#Poor-persons standardization\n",
    "\n",
    "print(np.mean(pT_true/100-2))\n",
    "print(np.std(pT_true/100-2))\n",
    "\n",
    "print(np.mean(m_true/10-2))\n",
    "print(np.std(m_true/10-2))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "e6f4cf12",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "ITERATION: 1\n",
      "STEP 1...\n",
      "Epoch 1/1000\n",
      "WARNING:tensorflow:From /global/home/users/bpnachman/anaconda3/envs/tf2_a40/lib/python3.8/site-packages/tensorflow/python/ops/array_ops.py:5043: calling gather (from tensorflow.python.ops.array_ops) with validate_indices is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "The `validate_indices` argument has no effect. Indices are always validated on CPU and never validated on GPU.\n",
      "7500/7500 [==============================] - 23s 3ms/step - loss: 0.6850 - accuracy: 0.5050 - val_loss: 0.6848 - val_accuracy: 0.4955\n",
      "Epoch 2/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6848 - accuracy: 0.5122 - val_loss: 0.6847 - val_accuracy: 0.5156\n",
      "Epoch 3/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6847 - accuracy: 0.5094 - val_loss: 0.6847 - val_accuracy: 0.5248\n",
      "Epoch 4/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6847 - accuracy: 0.5094 - val_loss: 0.6846 - val_accuracy: 0.5197\n",
      "Epoch 5/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6847 - accuracy: 0.5103 - val_loss: 0.6848 - val_accuracy: 0.4952\n",
      "Epoch 6/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6847 - accuracy: 0.5116 - val_loss: 0.6846 - val_accuracy: 0.5018\n",
      "Epoch 7/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6846 - accuracy: 0.5131 - val_loss: 0.6846 - val_accuracy: 0.5131\n",
      "Epoch 8/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6846 - accuracy: 0.5092 - val_loss: 0.6847 - val_accuracy: 0.4885\n",
      "Epoch 9/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6846 - accuracy: 0.5090 - val_loss: 0.6846 - val_accuracy: 0.5016\n",
      "Epoch 10/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6846 - accuracy: 0.5123 - val_loss: 0.6846 - val_accuracy: 0.5127\n",
      "Epoch 11/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6846 - accuracy: 0.5115 - val_loss: 0.6846 - val_accuracy: 0.5237\n",
      "Epoch 12/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6846 - accuracy: 0.5125 - val_loss: 0.6847 - val_accuracy: 0.5048\n",
      "Epoch 13/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6846 - accuracy: 0.5129 - val_loss: 0.6846 - val_accuracy: 0.5110\n",
      "Epoch 14/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6846 - accuracy: 0.5106 - val_loss: 0.6847 - val_accuracy: 0.5376\n",
      "Epoch 15/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6846 - accuracy: 0.5134 - val_loss: 0.6847 - val_accuracy: 0.5090\n",
      "Epoch 16/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6846 - accuracy: 0.5129 - val_loss: 0.6846 - val_accuracy: 0.5163\n",
      "Epoch 17/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6846 - accuracy: 0.5115 - val_loss: 0.6846 - val_accuracy: 0.5130\n",
      "Epoch 18/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6846 - accuracy: 0.5134 - val_loss: 0.6847 - val_accuracy: 0.5109\n",
      "Epoch 19/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6846 - accuracy: 0.5126 - val_loss: 0.6847 - val_accuracy: 0.5248\n",
      "Epoch 20/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6846 - accuracy: 0.5113 - val_loss: 0.6846 - val_accuracy: 0.5192\n",
      "Epoch 21/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6846 - accuracy: 0.5130 - val_loss: 0.6846 - val_accuracy: 0.5211\n",
      "Epoch 22/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6846 - accuracy: 0.5127 - val_loss: 0.6847 - val_accuracy: 0.5240\n",
      "Epoch 23/1000\n",
      "7500/7500 [==============================] - 21s 3ms/step - loss: 0.6846 - accuracy: 0.5107 - val_loss: 0.6846 - val_accuracy: 0.5160\n",
      "Restoring model weights from the end of the best epoch.\n",
      "Epoch 00023: early stopping\n",
      "STEP 2...\n",
      "Epoch 1/1000\n",
      "7500/7500 [==============================] - 21s 3ms/step - loss: 0.6866 - accuracy: 0.3442 - val_loss: 0.6863 - val_accuracy: 0.3344\n",
      "Epoch 2/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6866 - accuracy: 0.3410 - val_loss: 0.6863 - val_accuracy: 0.3497\n",
      "Epoch 3/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6866 - accuracy: 0.3412 - val_loss: 0.6863 - val_accuracy: 0.3464\n",
      "Epoch 4/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6866 - accuracy: 0.3426 - val_loss: 0.6863 - val_accuracy: 0.3504\n",
      "Epoch 5/1000\n",
      "7500/7500 [==============================] - 21s 3ms/step - loss: 0.6866 - accuracy: 0.3413 - val_loss: 0.6864 - val_accuracy: 0.3291\n",
      "Epoch 6/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6866 - accuracy: 0.3421 - val_loss: 0.6863 - val_accuracy: 0.3384\n",
      "Epoch 7/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6866 - accuracy: 0.3411 - val_loss: 0.6863 - val_accuracy: 0.3403\n",
      "Epoch 8/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6865 - accuracy: 0.3409 - val_loss: 0.6863 - val_accuracy: 0.3450\n",
      "Epoch 9/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6866 - accuracy: 0.3404 - val_loss: 0.6862 - val_accuracy: 0.3442\n",
      "Epoch 10/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6865 - accuracy: 0.3412 - val_loss: 0.6862 - val_accuracy: 0.3397\n",
      "Epoch 11/1000\n",
      "7500/7500 [==============================] - 21s 3ms/step - loss: 0.6865 - accuracy: 0.3410 - val_loss: 0.6863 - val_accuracy: 0.3478\n",
      "Epoch 12/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6865 - accuracy: 0.3400 - val_loss: 0.6863 - val_accuracy: 0.3475\n",
      "Epoch 13/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6865 - accuracy: 0.3410 - val_loss: 0.6863 - val_accuracy: 0.3434\n",
      "Epoch 14/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6865 - accuracy: 0.3405 - val_loss: 0.6865 - val_accuracy: 0.3562\n",
      "Epoch 15/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6865 - accuracy: 0.3404 - val_loss: 0.6864 - val_accuracy: 0.3534\n",
      "Epoch 16/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6865 - accuracy: 0.3406 - val_loss: 0.6863 - val_accuracy: 0.3398\n",
      "Epoch 17/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6865 - accuracy: 0.3400 - val_loss: 0.6863 - val_accuracy: 0.3423\n",
      "Epoch 18/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6865 - accuracy: 0.3412 - val_loss: 0.6862 - val_accuracy: 0.3375\n",
      "Epoch 19/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6865 - accuracy: 0.3399 - val_loss: 0.6863 - val_accuracy: 0.3350\n",
      "Epoch 20/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6865 - accuracy: 0.3386 - val_loss: 0.6862 - val_accuracy: 0.3394\n",
      "Restoring model weights from the end of the best epoch.\n",
      "Epoch 00020: early stopping\n",
      "ITERATION: 2\n",
      "STEP 1...\n",
      "Epoch 1/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6914 - accuracy: 0.4215 - val_loss: 0.6914 - val_accuracy: 0.4267\n",
      "Epoch 2/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6913 - accuracy: 0.4192 - val_loss: 0.6914 - val_accuracy: 0.4323\n",
      "Epoch 3/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6913 - accuracy: 0.4184 - val_loss: 0.6914 - val_accuracy: 0.4346\n",
      "Epoch 4/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6913 - accuracy: 0.4138 - val_loss: 0.6914 - val_accuracy: 0.4132\n",
      "Epoch 5/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6913 - accuracy: 0.4140 - val_loss: 0.6913 - val_accuracy: 0.4160\n",
      "Epoch 6/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6913 - accuracy: 0.4113 - val_loss: 0.6913 - val_accuracy: 0.4102\n",
      "Epoch 7/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6913 - accuracy: 0.4101 - val_loss: 0.6913 - val_accuracy: 0.3954\n",
      "Epoch 8/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6913 - accuracy: 0.4105 - val_loss: 0.6913 - val_accuracy: 0.4113\n",
      "Epoch 9/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6913 - accuracy: 0.4115 - val_loss: 0.6913 - val_accuracy: 0.4101\n",
      "Epoch 10/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6913 - accuracy: 0.4105 - val_loss: 0.6913 - val_accuracy: 0.4102\n",
      "Epoch 11/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6913 - accuracy: 0.4098 - val_loss: 0.6913 - val_accuracy: 0.4230\n",
      "Epoch 12/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6913 - accuracy: 0.4110 - val_loss: 0.6913 - val_accuracy: 0.4128\n",
      "Epoch 13/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6913 - accuracy: 0.4089 - val_loss: 0.6913 - val_accuracy: 0.4138\n",
      "Epoch 14/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6913 - accuracy: 0.4097 - val_loss: 0.6913 - val_accuracy: 0.4113\n",
      "Epoch 15/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6913 - accuracy: 0.4105 - val_loss: 0.6913 - val_accuracy: 0.4103\n",
      "Epoch 16/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6913 - accuracy: 0.4085 - val_loss: 0.6913 - val_accuracy: 0.4035\n",
      "Epoch 17/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6912 - accuracy: 0.4095 - val_loss: 0.6913 - val_accuracy: 0.4090\n",
      "Epoch 18/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6913 - accuracy: 0.4056 - val_loss: 0.6913 - val_accuracy: 0.4119\n",
      "Epoch 19/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6913 - accuracy: 0.4099 - val_loss: 0.6913 - val_accuracy: 0.4009\n",
      "Epoch 20/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6912 - accuracy: 0.4116 - val_loss: 0.6913 - val_accuracy: 0.4184\n",
      "Epoch 21/1000\n",
      "7500/7500 [==============================] - 21s 3ms/step - loss: 0.6912 - accuracy: 0.4090 - val_loss: 0.6913 - val_accuracy: 0.4195\n",
      "Epoch 22/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6913 - accuracy: 0.4094 - val_loss: 0.6914 - val_accuracy: 0.4240\n",
      "Epoch 23/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6912 - accuracy: 0.4112 - val_loss: 0.6913 - val_accuracy: 0.4045\n",
      "Epoch 24/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6912 - accuracy: 0.4090 - val_loss: 0.6915 - val_accuracy: 0.4295\n",
      "Epoch 25/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6913 - accuracy: 0.4097 - val_loss: 0.6913 - val_accuracy: 0.4196\n",
      "Epoch 26/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6912 - accuracy: 0.4102 - val_loss: 0.6913 - val_accuracy: 0.4125\n",
      "Epoch 27/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6912 - accuracy: 0.4105 - val_loss: 0.6912 - val_accuracy: 0.4113\n",
      "Epoch 28/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6912 - accuracy: 0.4096 - val_loss: 0.6913 - val_accuracy: 0.4163\n",
      "Epoch 29/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6913 - accuracy: 0.4101 - val_loss: 0.6913 - val_accuracy: 0.4174\n",
      "Epoch 30/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6912 - accuracy: 0.4116 - val_loss: 0.6914 - val_accuracy: 0.4285\n",
      "Epoch 31/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6912 - accuracy: 0.4107 - val_loss: 0.6913 - val_accuracy: 0.4173\n",
      "Epoch 32/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6912 - accuracy: 0.4128 - val_loss: 0.6912 - val_accuracy: 0.4131\n",
      "Epoch 33/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6912 - accuracy: 0.4098 - val_loss: 0.6913 - val_accuracy: 0.4136\n",
      "Epoch 34/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6912 - accuracy: 0.4118 - val_loss: 0.6912 - val_accuracy: 0.4143\n",
      "Epoch 35/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6912 - accuracy: 0.4125 - val_loss: 0.6913 - val_accuracy: 0.4102\n",
      "Epoch 36/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6912 - accuracy: 0.4101 - val_loss: 0.6913 - val_accuracy: 0.4168\n",
      "Epoch 37/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6912 - accuracy: 0.4113 - val_loss: 0.6913 - val_accuracy: 0.4068\n",
      "Epoch 38/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6912 - accuracy: 0.4072 - val_loss: 0.6913 - val_accuracy: 0.4165\n",
      "Epoch 39/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6912 - accuracy: 0.4125 - val_loss: 0.6914 - val_accuracy: 0.4255\n",
      "Epoch 40/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6912 - accuracy: 0.4107 - val_loss: 0.6913 - val_accuracy: 0.4236\n",
      "Epoch 41/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6912 - accuracy: 0.4099 - val_loss: 0.6913 - val_accuracy: 0.4128\n",
      "Epoch 42/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6912 - accuracy: 0.4088 - val_loss: 0.6913 - val_accuracy: 0.4185\n",
      "Epoch 43/1000\n",
      "7500/7500 [==============================] - 20s 3ms/step - loss: 0.6912 - accuracy: 0.4108 - val_loss: 0.6913 - val_accuracy: 0.4111\n",
      "Epoch 44/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6912 - accuracy: 0.4083 - val_loss: 0.6913 - val_accuracy: 0.4159\n",
      "Restoring model weights from the end of the best epoch.\n",
      "Epoch 00044: early stopping\n",
      "STEP 2...\n",
      "Epoch 1/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6831 - accuracy: 0.3513 - val_loss: 0.6830 - val_accuracy: 0.3551\n",
      "Epoch 2/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6828 - accuracy: 0.3461 - val_loss: 0.6829 - val_accuracy: 0.3511\n",
      "Epoch 3/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6828 - accuracy: 0.3445 - val_loss: 0.6830 - val_accuracy: 0.3566\n",
      "Epoch 4/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6828 - accuracy: 0.3454 - val_loss: 0.6830 - val_accuracy: 0.3444\n",
      "Epoch 5/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6828 - accuracy: 0.3466 - val_loss: 0.6829 - val_accuracy: 0.3538\n",
      "Epoch 6/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6828 - accuracy: 0.3469 - val_loss: 0.6830 - val_accuracy: 0.3407\n",
      "Epoch 7/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6828 - accuracy: 0.3474 - val_loss: 0.6829 - val_accuracy: 0.3524\n",
      "Epoch 8/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6828 - accuracy: 0.3476 - val_loss: 0.6829 - val_accuracy: 0.3432\n",
      "Epoch 9/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6828 - accuracy: 0.3469 - val_loss: 0.6830 - val_accuracy: 0.3572\n",
      "Epoch 10/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6828 - accuracy: 0.3470 - val_loss: 0.6830 - val_accuracy: 0.3583\n",
      "Epoch 11/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6828 - accuracy: 0.3472 - val_loss: 0.6830 - val_accuracy: 0.3541\n",
      "Epoch 12/1000\n",
      "7500/7500 [==============================] - 22s 3ms/step - loss: 0.6828 - accuracy: 0.3463 - val_loss: 0.6829 - val_accuracy: 0.3478\n",
      "Epoch 13/1000\n",
      "3343/7500 [============>.................] - ETA: 10s - loss: 0.6829 - accuracy: 0.3463"
     ]
    }
   ],
   "source": [
    "weights, models, history = multifold(dims, iterations, np.c_[pT_true/100.-2,m_true/10.-2], np.c_[pT_reco/100.-2,m_reco/10.-2], np.c_[pT_reco_alt/100.-2,m_reco_alt/10.-2], verbose=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "05a0ab8b",
   "metadata": {},
   "outputs": [],
   "source": [
    "for i in range(iterations):\n",
    "    fig = plt.figure(figsize=(8, 6)) \n",
    "    gs = gridspec.GridSpec(1, 1, height_ratios=[1]) \n",
    "    ax0 = plt.subplot(gs[0])\n",
    "    ax0.yaxis.set_ticks_position('both')\n",
    "    ax0.xaxis.set_ticks_position('both')\n",
    "    ax0.tick_params(direction=\"in\",which=\"both\")\n",
    "    ax0.minorticks_on()\n",
    "    plt.xticks(fontsize=20)\n",
    "    plt.yticks(fontsize=20)\n",
    "\n",
    "    OF,_,_  = plt.hist(m_true_alt,bins=np.linspace(0,70,30),color=\"blue\",label=\"Target\",density=True,alpha=0.2)\n",
    "    T,_,_  = plt.hist(m_true,bins=np.linspace(0,70,30),ls=\":\",histtype=\"step\",color=\"black\",lw=3,label=\"Prior\",density=True)\n",
    "    G,_,_ = plt.hist(m_true,weights=weights[i,1],bins=np.linspace(0,70,30),lw=3,histtype=\"step\",ls=\"-\",color=\"black\",label=\"OmniFold (it. \"+str(i)+\")\",density=True)\n",
    "    plt.legend()\n",
    "    plt.xlabel(\"Truth jet mass [GeV]\")\n",
    "    plt.ylabel(\"Normalized to unity\")\n",
    "\n",
    "    fig.savefig('figures/OmniFold_iteration_'+str(i)+'.pdf',bbox_inches='tight')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "06709fa7",
   "metadata": {},
   "outputs": [],
   "source": [
    "for i in range(iterations):\n",
    "    fig = plt.figure(figsize=(8, 6)) \n",
    "    gs = gridspec.GridSpec(1, 1, height_ratios=[1]) \n",
    "    ax0 = plt.subplot(gs[0])\n",
    "    ax0.yaxis.set_ticks_position('both')\n",
    "    ax0.xaxis.set_ticks_position('both')\n",
    "    ax0.tick_params(direction=\"in\",which=\"both\")\n",
    "    ax0.minorticks_on()\n",
    "    plt.xticks(fontsize=20)\n",
    "    plt.yticks(fontsize=20)\n",
    "\n",
    "    OF,_,_  = plt.hist(m_reco_alt,bins=np.linspace(0,70,30),color=\"blue\",label=\"Target\",density=True,alpha=0.2)\n",
    "    T,_,_  = plt.hist(m_reco,bins=np.linspace(0,70,30),ls=\":\",histtype=\"step\",color=\"black\",lw=3,label=\"Prior\",density=True)\n",
    "    G,_,_ = plt.hist(m_reco,weights=weights[i,0],bins=np.linspace(0,70,30),lw=3,histtype=\"step\",ls=\"-\",color=\"black\",label=\"OmniFold (it. \"+str(i)+\")\",density=True)\n",
    "    plt.legend()\n",
    "    plt.xlabel(\"Reconstructed jet mass [GeV]\")\n",
    "    plt.ylabel(\"Normalized to unity\")\n",
    "\n",
    "    fig.savefig('figures/OmniFold_iteration_'+str(i)+'_reco.pdf',bbox_inches='tight')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "10fee4a8",
   "metadata": {},
   "outputs": [],
   "source": [
    "plt.hist(m_reco,weights=weights[-1,1],bins=np.linspace(0,50,20),histtype=\"step\",ls=\"-\",color=\"black\",label=\"rw\")\n",
    "plt.hist(m_reco_alt,bins=np.linspace(0,50,20),ls=\":\",histtype=\"step\",color=\"black\",lw=3,label=\"target\")\n",
    "plt.hist(m_reco,bins=np.linspace(0,50,20),alpha=0.2,color=\"blue\",label=\"prior\")\n",
    "plt.legend()\n",
    "plt.xlabel(\"detector-level jet mass\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "895d3e89",
   "metadata": {},
   "outputs": [],
   "source": [
    "plt.hist(m_reco,weights=weights[-1,0],bins=np.linspace(0,50,20),histtype=\"step\",ls=\"-\",color=\"black\",label=\"rw\")\n",
    "plt.hist(m_reco_alt,bins=np.linspace(0,50,20),ls=\":\",histtype=\"step\",color=\"black\",lw=3,label=\"target\")\n",
    "plt.hist(m_reco,bins=np.linspace(0,50,20),alpha=0.2,color=\"blue\",label=\"prior\")\n",
    "plt.legend()\n",
    "plt.xlabel(\"detector-level jet mass\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "e73a532c",
   "metadata": {},
   "outputs": [],
   "source": [
    "#poor-person standardization\n",
    "\n",
    "print(np.mean(w_true*10-1))\n",
    "print(np.std(w_true*10-1))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "fd5e9f5f",
   "metadata": {},
   "outputs": [],
   "source": [
    "weightsW, modelsW, historyW = multifold(dims, iterations, np.c_[pT_true/100.-2,w_true*10.-1], np.c_[pT_reco/100.-2,w_reco*10.-1], np.c_[pT_reco_alt/100.-2,w_reco_alt*10.-1], verbose=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "8b43b25c",
   "metadata": {},
   "outputs": [],
   "source": [
    "for i in range(iterations):\n",
    "    fig = plt.figure(figsize=(8, 6)) \n",
    "    gs = gridspec.GridSpec(1, 1, height_ratios=[1]) \n",
    "    ax0 = plt.subplot(gs[0])\n",
    "    ax0.yaxis.set_ticks_position('both')\n",
    "    ax0.xaxis.set_ticks_position('both')\n",
    "    ax0.tick_params(direction=\"in\",which=\"both\")\n",
    "    ax0.minorticks_on()\n",
    "    plt.xticks(fontsize=20)\n",
    "    plt.yticks(fontsize=20)\n",
    "\n",
    "    OF,_,_  = plt.hist(w_reco_alt,bins=np.linspace(0,0.6,30),color=\"blue\",label=\"Target\",density=True,alpha=0.2)\n",
    "    T,_,_  = plt.hist(w_reco,bins=np.linspace(0,0.6,30),ls=\":\",histtype=\"step\",color=\"black\",lw=3,label=\"Prior\",density=True)\n",
    "    G,_,_ = plt.hist(w_reco,weights=weightsW[i,0],bins=np.linspace(0,0.6,30),lw=3,histtype=\"step\",ls=\"-\",color=\"black\",label=\"OmniFold (it. \"+str(i)+\")\",density=True)\n",
    "    plt.legend()\n",
    "    plt.xlabel(\"Reconstructed jet width\")\n",
    "    plt.ylabel(\"Normalized to unity\")\n",
    "\n",
    "    fig.savefig('figures/OmniFoldW_iteration_'+str(i)+'_reco.pdf',bbox_inches='tight')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "e3899254",
   "metadata": {},
   "outputs": [],
   "source": [
    "for i in range(iterations):\n",
    "    fig = plt.figure(figsize=(8, 6)) \n",
    "    gs = gridspec.GridSpec(1, 1, height_ratios=[1]) \n",
    "    ax0 = plt.subplot(gs[0])\n",
    "    ax0.yaxis.set_ticks_position('both')\n",
    "    ax0.xaxis.set_ticks_position('both')\n",
    "    ax0.tick_params(direction=\"in\",which=\"both\")\n",
    "    ax0.minorticks_on()\n",
    "    plt.xticks(fontsize=20)\n",
    "    plt.yticks(fontsize=20)\n",
    "\n",
    "    OF,_,_  = plt.hist(w_true_alt,bins=np.linspace(0,0.6,30),color=\"blue\",label=\"Target\",density=True,alpha=0.2)\n",
    "    T,_,_  = plt.hist(w_true,bins=np.linspace(0,0.6,30),ls=\":\",histtype=\"step\",color=\"black\",lw=3,label=\"Prior\",density=True)\n",
    "    G,_,_ = plt.hist(w_true,weights=weightsW[i,1],bins=np.linspace(0,0.6,30),lw=3,histtype=\"step\",ls=\"-\",color=\"black\",label=\"OmniFold (it. \"+str(i)+\")\",density=True)\n",
    "    plt.legend()\n",
    "    plt.xlabel(\"Truth jet width\")\n",
    "    plt.ylabel(\"Normalized to unity\")\n",
    "\n",
    "    fig.savefig('figures/OmniFoldW_iteration_'+str(i)+'.pdf',bbox_inches='tight')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "2f5be1ff",
   "metadata": {},
   "outputs": [],
   "source": [
    "GW,_,_ = plt.hist(w_true,weights=weightsW[-1,1],bins=np.linspace(0,1,30),histtype=\"step\",ls=\"-\",color=\"black\",label=\"rw\")\n",
    "OFW,_,_  = plt.hist(w_true_alt,bins=np.linspace(0,1,30),ls=\":\",histtype=\"step\",color=\"black\",lw=3,label=\"target\")\n",
    "TW,_,_  = plt.hist(w_true,bins=np.linspace(0,1,30),alpha=0.2,color=\"blue\",label=\"prior\")\n",
    "plt.legend()\n",
    "plt.xlabel(\"jet width\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "12dd28a3",
   "metadata": {},
   "outputs": [],
   "source": [
    "#poor-person standardization\n",
    "\n",
    "print(np.mean(np.array(q_true)*10))\n",
    "print(np.std(np.array(q_true)*10))\n",
    "\n",
    "print(np.mean(np.array(r_true)*5-3))\n",
    "print(np.std(np.array(r_true)*5-3))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "53136cb2",
   "metadata": {},
   "outputs": [],
   "source": [
    "weightsQ, modelsQ, historyQ = multifold(dims, iterations, np.c_[pT_true/100.-2,q_true*10.], np.c_[pT_reco/100.-2,q_reco*10.], np.c_[pT_reco_alt/100.-2,q_reco_alt*10.], verbose=1)\n",
    "weightsR, modelsR, historyR = multifold(dims, iterations, np.c_[pT_true/100.-2,r_true*5.-3], np.c_[pT_reco/100.-2,r_reco*5.-3], np.c_[pT_reco_alt/100.-2,r_reco_alt*5.-3], verbose=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "8a1e1873",
   "metadata": {},
   "outputs": [],
   "source": [
    "_,_,_ = plt.hist(q_true,weights=weightsQ[-1,1],bins=np.linspace(-0.5,0.5,20),histtype=\"step\",ls=\"-\",color=\"black\",label=\"rw\")\n",
    "_,_,_  = plt.hist(q_true_alt,bins=np.linspace(-0.5,0.5,20),ls=\":\",histtype=\"step\",color=\"black\",lw=3,label=\"target\")\n",
    "_,_,_  = plt.hist(q_true,bins=np.linspace(-0.5,0.5,20),alpha=0.2,color=\"blue\",label=\"prior\")\n",
    "plt.legend()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "d04a23b5",
   "metadata": {},
   "outputs": [],
   "source": [
    "_,_,_ = plt.hist(r_true,weights=weightsR[-1,1],bins=np.linspace(0,1.5,20),histtype=\"step\",ls=\"-\",color=\"black\",label=\"rw\")\n",
    "_,_,_  = plt.hist(r_true_alt,bins=np.linspace(0,1.5,20),ls=\":\",histtype=\"step\",color=\"black\",lw=3,label=\"target\")\n",
    "_,_,_  = plt.hist(r_true,bins=np.linspace(0,1.5,20),alpha=0.2,color=\"blue\",label=\"prior\")\n",
    "plt.legend()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "cbb32a59",
   "metadata": {
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "for i in range(iterations):\n",
    "    fig = plt.figure(figsize=(8, 6)) \n",
    "    gs = gridspec.GridSpec(1, 1, height_ratios=[1]) \n",
    "    ax0 = plt.subplot(gs[0])\n",
    "    ax0.yaxis.set_ticks_position('both')\n",
    "    ax0.xaxis.set_ticks_position('both')\n",
    "    ax0.tick_params(direction=\"in\",which=\"both\")\n",
    "    ax0.minorticks_on()\n",
    "    plt.xticks(fontsize=20)\n",
    "    plt.yticks(fontsize=20)\n",
    "\n",
    "    OF,_,_  = plt.hist(q_true_alt,bins=np.linspace(-0.5,0.5,30),color=\"blue\",label=\"Target\",density=True,alpha=0.2)\n",
    "    T,_,_  = plt.hist(q_true,bins=np.linspace(-0.5,0.5,30),ls=\":\",histtype=\"step\",color=\"black\",lw=3,label=\"Prior\",density=True)\n",
    "    G,_,_ = plt.hist(q_true,weights=weightsQ[i,1],bins=np.linspace(-0.5,0.5,30),lw=3,histtype=\"step\",ls=\"-\",color=\"black\",label=\"OmniFold (it. \"+str(i)+\")\",density=True)\n",
    "    plt.legend()\n",
    "    plt.xlabel(\"Truth jet charge\")\n",
    "    plt.ylabel(\"Normalized to unity\")\n",
    "\n",
    "    fig.savefig('figures/OmniFoldQ_iteration_'+str(i)+'.pdf',bbox_inches='tight')\n",
    "    \n",
    "for i in range(iterations):\n",
    "    fig = plt.figure(figsize=(8, 6)) \n",
    "    gs = gridspec.GridSpec(1, 1, height_ratios=[1]) \n",
    "    ax0 = plt.subplot(gs[0])\n",
    "    ax0.yaxis.set_ticks_position('both')\n",
    "    ax0.xaxis.set_ticks_position('both')\n",
    "    ax0.tick_params(direction=\"in\",which=\"both\")\n",
    "    ax0.minorticks_on()\n",
    "    plt.xticks(fontsize=20)\n",
    "    plt.yticks(fontsize=20)\n",
    "\n",
    "    OF,_,_  = plt.hist(q_reco_alt,bins=np.linspace(-0.5,0.5,30),color=\"blue\",label=\"Target\",density=True,alpha=0.2)\n",
    "    T,_,_  = plt.hist(q_reco,bins=np.linspace(-0.5,0.5,30),ls=\":\",histtype=\"step\",color=\"black\",lw=3,label=\"Prior\",density=True)\n",
    "    G,_,_ = plt.hist(q_reco,weights=weightsQ[i,0],bins=np.linspace(-0.5,0.5,30),lw=3,histtype=\"step\",ls=\"-\",color=\"black\",label=\"OmniFold (it. \"+str(i)+\")\",density=True)\n",
    "    plt.legend()\n",
    "    plt.xlabel(\"Reconstructed jet charge\")\n",
    "    plt.ylabel(\"Normalized to unity\")\n",
    "\n",
    "    fig.savefig('figures/OmniFoldQ_iteration_'+str(i)+'_reco.pdf',bbox_inches='tight')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "676e2b9f",
   "metadata": {
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "for i in range(iterations):\n",
    "    fig = plt.figure(figsize=(8, 6)) \n",
    "    gs = gridspec.GridSpec(1, 1, height_ratios=[1]) \n",
    "    ax0 = plt.subplot(gs[0])\n",
    "    ax0.yaxis.set_ticks_position('both')\n",
    "    ax0.xaxis.set_ticks_position('both')\n",
    "    ax0.tick_params(direction=\"in\",which=\"both\")\n",
    "    ax0.minorticks_on()\n",
    "    plt.xticks(fontsize=20)\n",
    "    plt.yticks(fontsize=20)\n",
    "\n",
    "    OF,_,_  = plt.hist(r_true_alt,bins=np.linspace(0,1,30),color=\"blue\",label=\"Target\",density=True,alpha=0.2)\n",
    "    T,_,_  = plt.hist(r_true,bins=np.linspace(0,1,30),ls=\":\",histtype=\"step\",color=\"black\",lw=3,label=\"Prior\",density=True)\n",
    "    G,_,_ = plt.hist(r_true,weights=weightsR[i,1],bins=np.linspace(0,1,30),lw=3,histtype=\"step\",ls=\"-\",color=\"black\",label=\"OmniFold (it. \"+str(i)+\")\",density=True)\n",
    "    plt.legend()\n",
    "    plt.xlabel(\"Truth jet charge fraction\")\n",
    "    plt.ylabel(\"Normalized to unity\")\n",
    "\n",
    "    fig.savefig('figures/OmniFoldR_iteration_'+str(i)+'.pdf',bbox_inches='tight')\n",
    "    \n",
    "for i in range(iterations):\n",
    "    fig = plt.figure(figsize=(8, 6)) \n",
    "    gs = gridspec.GridSpec(1, 1, height_ratios=[1]) \n",
    "    ax0 = plt.subplot(gs[0])\n",
    "    ax0.yaxis.set_ticks_position('both')\n",
    "    ax0.xaxis.set_ticks_position('both')\n",
    "    ax0.tick_params(direction=\"in\",which=\"both\")\n",
    "    ax0.minorticks_on()\n",
    "    plt.xticks(fontsize=20)\n",
    "    plt.yticks(fontsize=20)\n",
    "\n",
    "    OF,_,_  = plt.hist(r_reco_alt,bins=np.linspace(0,1,30),color=\"blue\",label=\"Target\",density=True,alpha=0.2)\n",
    "    T,_,_  = plt.hist(r_reco,bins=np.linspace(0,1,30),ls=\":\",histtype=\"step\",color=\"black\",lw=3,label=\"Prior\",density=True)\n",
    "    G,_,_ = plt.hist(r_reco,weights=weightsR[i,0],bins=np.linspace(0,1,30),lw=3,histtype=\"step\",ls=\"-\",color=\"black\",label=\"OmniFold (it. \"+str(i)+\")\",density=True)\n",
    "    plt.legend()\n",
    "    plt.xlabel(\"Reconstructed jet charge fraction\")\n",
    "    plt.ylabel(\"Normalized to unity\")\n",
    "\n",
    "    fig.savefig('figures/OmniFoldR_iteration_'+str(i)+'_reco.pdf',bbox_inches='tight')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "e7873829",
   "metadata": {},
   "outputs": [],
   "source": [
    "pT_true = datasets['Pythia26']['gen_jets'][:,0]\n",
    "pT_reco = datasets['Pythia26']['sim_jets'][:,0]\n",
    "\n",
    "pT_true_alt = datasets['Herwig']['gen_jets'][:,0]\n",
    "pT_reco_alt = datasets['Herwig']['sim_jets'][:,0]\n",
    "\n",
    "features = {}\n",
    "\n",
    "features[\"m\",\"rec\",\"nom\"] = np.array(datasets['Pythia26']['sim_jets'][:,3])\n",
    "features[\"m\",\"tru\",\"nom\"] = np.array(datasets['Pythia26']['gen_jets'][:,3])\n",
    "features[\"m\",\"rec\",\"alt\"] = np.array(datasets['Herwig']['sim_jets'][:,3])\n",
    "features[\"m\",\"tru\",\"alt\"] = np.array(datasets['Herwig']['gen_jets'][:,3])\n",
    "\n",
    "features[\"w\",\"rec\",\"nom\"] = np.array(datasets['Pythia26']['sim_widths'])\n",
    "features[\"w\",\"tru\",\"nom\"] = np.array(datasets['Pythia26']['gen_widths'])\n",
    "features[\"w\",\"rec\",\"alt\"] = np.array(datasets['Herwig']['sim_widths'])\n",
    "features[\"w\",\"tru\",\"alt\"] = np.array(datasets['Herwig']['gen_widths'])\n",
    "\n",
    "features[\"q\",\"rec\",\"nom\"] = np.array(datasets['Pythia26']['sim_charge'])\n",
    "features[\"q\",\"tru\",\"nom\"] = np.array(datasets['Pythia26']['gen_charge'])\n",
    "features[\"q\",\"rec\",\"alt\"] = np.array(datasets['Herwig']['sim_charge'])\n",
    "features[\"q\",\"tru\",\"alt\"] = np.array(datasets['Herwig']['gen_charge'])\n",
    "\n",
    "features[\"r\",\"rec\",\"nom\"] = np.array(datasets['Pythia26']['sim_pTcharge'])\n",
    "features[\"r\",\"tru\",\"nom\"] = np.array(datasets['Pythia26']['gen_pTcharge'])\n",
    "features[\"r\",\"rec\",\"alt\"] = np.array(datasets['Herwig']['sim_pTcharge'])\n",
    "features[\"r\",\"tru\",\"alt\"] = np.array(datasets['Herwig']['gen_pTcharge'])\n",
    "\n",
    "maxvalues = {}\n",
    "\n",
    "maxvalues['m'] = 200\n",
    "maxvalues['w'] = 0.7\n",
    "maxvalues['q'] = 0.5\n",
    "maxvalues['r'] = 1\n",
    "\n",
    "minvalues = {}\n",
    "\n",
    "minvalues['m'] = 0\n",
    "minvalues['w'] = 0\n",
    "minvalues['q'] = -0.5\n",
    "minvalues['r'] = 0\n",
    "\n",
    "omnifolded = {}\n",
    "omnifolded['r'] = weightsR[-1,1]\n",
    "omnifolded['q'] = weightsQ[-1,1]\n",
    "omnifolded['w'] = weightsW[-1,1]\n",
    "omnifolded['m'] = weights[-1,1]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "f5a1d393",
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "binvalsObs = {}\n",
    "for obs in ['r','q','w','m']:\n",
    "    binvalsObs[obs] = {}\n",
    "    pTbin_truth = np.clip(np.digitize(pT_true,binvals),1,len(binvals)-1)-1\n",
    "    pTbin_reco = np.clip(np.digitize(pT_reco,binvals),1,len(binvals)-1)-1\n",
    "    xt = features[obs,\"tru\",\"nom\"]\n",
    "    xr = features[obs,\"rec\",\"nom\"]\n",
    "    for ii in range(len(binvals)-1):\n",
    "        binvalsObs[obs][ii] = [minvalues[obs]]\n",
    "        i = 0\n",
    "        disttotal = 0.\n",
    "        breakloop = True\n",
    "        while len(binvalsObs[obs][ii]) < 15 and binvalsObs[obs][ii][-1] < maxvalues[obs] and breakloop:\n",
    "            mycount = 0\n",
    "            for binhigh in np.linspace(binvalsObs[obs][ii][i]+0.01,maxvalues[obs],100):\n",
    "                mycount+=1\n",
    "                purity = len(xt[(pTbin_reco==ii)*(pTbin_truth==ii)*(xt > binvalsObs[obs][ii][i])*(xt < binhigh)*(xr > binvalsObs[obs][ii][i])*(xr < binhigh)]) / (0.00000001+len(xt[(pTbin_truth==ii)*(xt > binvalsObs[obs][ii][i])*(xt < binhigh)]))\n",
    "                distamount = len(xt[(pTbin_reco==ii)*(pTbin_truth==ii)*(xr > binvalsObs[obs][ii][i])*(xr < binhigh)]) / (0.00000001+len(xr[(pTbin_reco==ii)*(pTbin_truth==ii)]))\n",
    "                if (purity > 0.5):\n",
    "                    i+=1\n",
    "                    disttotal += distamount\n",
    "                    binvalsObs[obs][ii]+=[binhigh]\n",
    "                    #print(\"  \",binhigh,purity,disttotal,distamount)\n",
    "                    break\n",
    "                if (mycount==99):\n",
    "                    breakloop = False\n",
    "                    #binvalsObs[obs][ii]+=[rmax]\n",
    "\n",
    "        print(len(binvalsObs[obs][ii])) #,binvalsObs[obs][ii])\n",
    "        pass\n",
    "    \n",
    "    #Next, let's construct the response matrix.\n",
    "\n",
    "    pTbin_truth = np.clip(np.digitize(pT_true,binvals),1,len(binvals)-1)-1\n",
    "    pTbin_reco = np.clip(np.digitize(pT_reco,binvals),1,len(binvals)-1)-1\n",
    "\n",
    "    xbin_truth_all = np.array([np.clip(np.digitize(xt,binvalsObs[obs][ii]),1,len(binvalsObs[obs][ii])-1)-1 for ii in range(len(binvals)-1)])\n",
    "    xbin_reco_all = np.array([np.clip(np.digitize(xr,binvalsObs[obs][ii]),1,len(binvalsObs[obs][ii])-1)-1 for ii in range(len(binvals)-1)] )\n",
    "    xbin_truth_all = xbin_truth_all.T\n",
    "    xbin_reco_all = xbin_reco_all.T\n",
    "\n",
    "    xbin_truth = np.array([xbin_truth_all[i][pTbin_truth[i]] for i in range(len(pTbin_truth))])\n",
    "    xbin_reco = np.array([xbin_reco_all[i][pTbin_reco[i]] for i in range(len(pTbin_reco))])\n",
    "\n",
    "    ###\n",
    "    bin2_truth = []\n",
    "    for i in range(len(pTbin_truth)):\n",
    "        mybin = 0\n",
    "        for i2 in range(0,pTbin_truth[i]):\n",
    "            mybin+=len(binvalsObs[obs][i2])-1\n",
    "        bin2_truth+=[mybin+xbin_truth[i]]\n",
    "\n",
    "    bin2_reco = []\n",
    "    for i in range(len(pTbin_reco)):\n",
    "        mybin = 0\n",
    "        for i2 in range(0,pTbin_reco[i]):\n",
    "            mybin+=len(binvalsObs[obs][i2])-1\n",
    "        bin2_reco+=[mybin+xbin_reco[i]]\n",
    "\n",
    "    #alt\n",
    "    xta = features[obs,\"tru\",\"alt\"]\n",
    "    xra = features[obs,\"rec\",\"alt\"]\n",
    "\n",
    "    pTbin_truth_alt = np.clip(np.digitize(pT_true_alt,binvals),1,len(binvals)-1)-1\n",
    "    pTbin_reco_alt = np.clip(np.digitize(pT_reco_alt,binvals),1,len(binvals)-1)-1\n",
    "\n",
    "    xbin_truth_all_alt = np.array([np.clip(np.digitize(xta,binvalsObs[obs][ii]),1,len(binvalsObs[obs][ii])-1)-1 for ii in range(len(binvals)-1)])\n",
    "    xbin_reco_all_alt = np.array([np.clip(np.digitize(xra,binvalsObs[obs][ii]),1,len(binvalsObs[obs][ii])-1)-1 for ii in range(len(binvals)-1)] )\n",
    "    xbin_truth_all_alt = xbin_truth_all_alt.T\n",
    "    xbin_reco_all_alt = xbin_reco_all_alt.T\n",
    "\n",
    "    xbin_truth_alt = np.array([xbin_truth_all_alt[i][pTbin_truth_alt[i]] for i in range(len(pTbin_truth_alt))])\n",
    "    xbin_reco_alt = np.array([xbin_reco_all_alt[i][pTbin_reco_alt[i]] for i in range(len(pTbin_reco_alt))])\n",
    "\n",
    "    bin2_truth_alt = []\n",
    "    for i in range(len(pTbin_truth_alt)):\n",
    "        mybin = 0\n",
    "        for i2 in range(0,pTbin_truth_alt[i]):\n",
    "            mybin+=len(binvalsObs[obs][i2])-1\n",
    "        bin2_truth_alt+=[mybin+xbin_truth_alt[i]]\n",
    "\n",
    "    bin2_reco_alt = []\n",
    "    for i in range(len(pTbin_reco_alt)):\n",
    "        mybin = 0\n",
    "        for i2 in range(0,pTbin_reco_alt[i]):\n",
    "            mybin+=len(binvalsObs[obs][i2])-1\n",
    "        bin2_reco_alt+=[mybin+xbin_reco_alt[i]]\n",
    "\n",
    "    fig = plt.figure(figsize=(8, 6)) \n",
    "    gs = gridspec.GridSpec(1, 1, height_ratios=[1]) \n",
    "    ax0 = plt.subplot(gs[0])\n",
    "    ax0.yaxis.set_ticks_position('both')\n",
    "    ax0.xaxis.set_ticks_position('both')\n",
    "    ax0.tick_params(direction=\"in\",which=\"both\")\n",
    "    ax0.minorticks_on()\n",
    "    plt.xticks(fontsize=20)\n",
    "    plt.yticks(fontsize=20)\n",
    "\n",
    "    H, xedges, yedges = np.histogram2d(bin2_truth,bin2_reco,bins=[range(max(bin2_truth)+2),range(max(bin2_truth)+2)])\n",
    "    H_alt, xedges, yedges = np.histogram2d(bin2_truth_alt,bin2_reco_alt,bins=[range(max(bin2_truth)+2),range(max(bin2_truth)+2)])\n",
    "    H_norm = H / H.sum(axis=1, keepdims=True)\n",
    "    plt.imshow(H_norm,origin='lower',cmap=\"Reds\",vmin = 0,vmax = 1)\n",
    "    cbar = plt.colorbar()\n",
    "    cbar.ax.set_ylabel('Pr(Detector | Particle)')\n",
    "\n",
    "    plt.xlabel(\"Particle-level \"+obs+\" and $p_{T}$ bin\",fontsize=20)\n",
    "    plt.ylabel(\"Detector-level \"+obs+\" and $p_{T}$ bin\",fontsize=20)\n",
    "    fig.savefig('figures/response'+obs+'.pdf',bbox_inches='tight')\n",
    "    \n",
    "    #IBU\n",
    "    T = np.sum(H,axis=1)\n",
    "    D = np.sum(H,axis=0)\n",
    "    D_alt = np.sum(H_alt,axis=0)\n",
    "    T_alt = np.sum(H_alt,axis=1)\n",
    "    ibu = IBU(T,D_alt,H_norm.T,10)\n",
    "    #ibu = IBU(T,D,H_norm.T,10)\n",
    "    \n",
    "    for moment in [1,2]:\n",
    "\n",
    "        #Unbinned\n",
    "        means_unbinnedx = np.array([np.mean(xt[pTbin_truth==i]**moment) for i in range(7)])\n",
    "        means_unbinnedx_alt = np.array([np.mean(xta[pTbin_truth_alt==i]**moment) for i in range(7)])\n",
    "\n",
    "        means_unbinnedx_reco = np.array([np.mean(xr[pTbin_reco==i]**moment) for i in range(7)])\n",
    "        means_unbinnedx_reco_alt = np.array([np.mean(xra[pTbin_reco_alt==i]**moment) for i in range(7)])\n",
    "        means_unbinnedx_omnifold_alt = np.array([np.average(xt[pTbin_truth==i]**moment,weights=omnifolded[obs][pTbin_truth==i]) for i in range(7)])\n",
    "\n",
    "        #Various corrections\n",
    "        means_binnedx = []\n",
    "        means_binnedx_alt = []\n",
    "        means_binnedx_alt_corrected = []\n",
    "        for i in range(7):\n",
    "\n",
    "            mybin = 0\n",
    "            for i2 in range(0,i):\n",
    "                mybin += len(binvalsObs[obs][i2])-1\n",
    "\n",
    "            ibu_i = [ibu[mybin+j] for j in range(len(binvalsObs[obs][i])-1)]\n",
    "\n",
    "            x_centers = 0.5*(np.array(binvalsObs[obs][i][0:-1])+np.array(binvalsObs[obs][i][1:]))\n",
    "            xvals = [len(xt[(pTbin_truth==i)*(xbin_truth==j)]) for j in range(len(binvalsObs[obs][i])-1)]\n",
    "            xvals_alt = [len(xta[(pTbin_truth_alt==i)*(xbin_truth_alt==j)]) for j in range(len(binvalsObs[obs][i])-1)]\n",
    "            means_binnedx += [np.sum(x_centers**moment*xvals)/np.sum(xvals)]\n",
    "            means_binnedx_alt += [np.sum(x_centers**moment*ibu_i)/np.sum(ibu_i)]\n",
    "\n",
    "            #Try a per-bin correction\n",
    "            xiavg = []\n",
    "            for j in range(len(binvalsObs[obs][i])-1):\n",
    "                xiavg+=[np.mean(xt[(pTbin_truth==i)*(xbin_truth==j)]**moment)]\n",
    "            xiavg = np.array(xiavg)\n",
    "            #print(np.mean(xt[(pTbin_truth==i)]),np.sum(xiavg*ibu_i)/np.sum(ibu_i))\n",
    "            means_binnedx_alt_corrected += [np.sum(xiavg*ibu_i)/np.sum(ibu_i)]\n",
    "\n",
    "        means_binnedx = np.array(means_binnedx)\n",
    "        means_binnedx_alt = np.array(means_binnedx_alt)\n",
    "        means_binnedx_alt_corrected = np.array(means_binnedx_alt_corrected)\n",
    "\n",
    "        fig = plt.figure(figsize=(8, 8))\n",
    "        gs = gridspec.GridSpec(2, 1, height_ratios=[3,4]) \n",
    "        ax0 = plt.subplot(gs[0])\n",
    "        ax0.yaxis.set_ticks_position('both')\n",
    "        ax0.xaxis.set_ticks_position('both')\n",
    "        ax0.tick_params(direction=\"in\",which=\"both\")\n",
    "        plt.xticks(fontsize=0)\n",
    "        plt.yticks(fontsize=20)\n",
    "        ax0.minorticks_on()\n",
    "\n",
    "        plt.plot(0.5*(np.array(binvals[0:-1])+np.array(binvals[1:])),means_unbinnedx,marker='^',ls=\"\",label=\"Unbinned Pythia\",color='red')\n",
    "        plt.plot(0.5*(np.array(binvals[0:-1])+np.array(binvals[1:])),means_unbinnedx_alt,marker='v',ls=\"\",label=\"Unbinned Herwig\",color='black')\n",
    "        #plt.ylim([0.05,0.25])\n",
    "\n",
    "        plt.legend(frameon=True,fontsize=14)\n",
    "        if (moment==1):\n",
    "            plt.ylabel(r'$< '+obs+' >$',fontsize=20)\n",
    "        else:\n",
    "            plt.ylabel(r'$< '+obs+'^'+str(moment)+' >$',fontsize=20)\n",
    "        if (obs=='m'):\n",
    "            if (moment==1):\n",
    "                plt.ylabel(r'$< '+obs+' >$ [GeV]',fontsize=20)\n",
    "            else:\n",
    "                plt.ylabel(r'$< '+obs+'^'+str(moment)+' >$ [GeV$^'+str(moment)+'$]',fontsize=20)\n",
    "            \n",
    "        ax1 = plt.subplot(gs[1])\n",
    "        ax1.yaxis.set_ticks_position('both')\n",
    "        ax1.xaxis.set_ticks_position('both')\n",
    "        ax1.tick_params(direction=\"in\",which=\"both\")\n",
    "        ax1.minorticks_on()\n",
    "\n",
    "        ax1.plot(0.5*(np.array(binvals[0:-1])+np.array(binvals[1:])),means_binnedx_alt_corrected/means_unbinnedx_alt,marker='s',ls=\"\",label=\"IBU + Bin Avg Correction\")\n",
    "        ax1.plot(0.5*(np.array(binvals[0:-1])+np.array(binvals[1:])),(means_unbinnedx_reco_alt*means_unbinnedx/means_unbinnedx_reco)/means_unbinnedx_alt,marker='o',ls=\"\",label=\"IBU + Bin Correction\")\n",
    "        ax1.plot(0.5*(np.array(binvals[0:-1])+np.array(binvals[1:])),(means_binnedx_alt*means_unbinnedx/means_binnedx)/means_unbinnedx_alt,marker='s',ls=\"\",label=\"Naive Correction\")\n",
    "        ax1.plot(0.5*(np.array(binvals[0:-1])+np.array(binvals[1:])),means_unbinnedx_omnifold_alt/means_unbinnedx_alt,marker='s',ls=\"\",label=\"OmniFold\")\n",
    "        ax1.set_xlabel(\"Jet $p_{T}$ [GeV]\",fontsize=20)\n",
    "        ax1.axhline(1.,ls=\":\",color=\"black\")\n",
    "        ax1.set_ylim([0.85,1.2])\n",
    "        ax1.legend(frameon=True,ncol=2,fontsize=14)\n",
    "        plt.ylabel('Ratio to \\n Unbinned Herwig',fontsize=20)\n",
    "        fig.savefig('figures/moment'+str(moment)+\"_\"+obs+'_v_pT.pdf',bbox_inches='tight')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "57f78825",
   "metadata": {},
   "source": [
    "## Now, experiment with Moment OmniFolding"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "cdcca443",
   "metadata": {},
   "outputs": [],
   "source": [
    "gauss_data = np.random.normal(0,1,100000)\n",
    "gauss_sim = np.random.normal(-0.5,1,100000)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "e4a1450a",
   "metadata": {},
   "outputs": [],
   "source": [
    "def weighted_binary_crossentropy(y_true, y_pred):\n",
    "    weights = tf.gather(y_true, [1], axis=1) # event weights\n",
    "    y_true = tf.gather(y_true, [0], axis=1) # actual y_true for loss\n",
    "    \n",
    "    # Clip the prediction value to prevent NaN's and Inf's\n",
    "    epsilon = K.epsilon()\n",
    "    y_pred = K.clip(y_pred, epsilon, 1. - epsilon)\n",
    "    t_loss = -weights * ((y_true) * K.log(y_pred) +\n",
    "                         (1 - y_true) * K.log(1 - y_pred))\n",
    "    return K.mean(t_loss)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "8fae5716",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/10\n",
      "WARNING:tensorflow:From /global/home/users/bpnachman/anaconda3/envs/tf2_a40/lib/python3.8/site-packages/tensorflow/python/ops/array_ops.py:5043: calling gather (from tensorflow.python.ops.array_ops) with validate_indices is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "The `validate_indices` argument has no effect. Indices are always validated on CPU and never validated on GPU.\n",
      "150/150 [==============================] - 1s 2ms/step - loss: 0.5127 - accuracy: 0.4621\n",
      "Epoch 2/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.4714 - accuracy: 0.4869\n",
      "Epoch 3/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.4714 - accuracy: 0.4872\n",
      "Epoch 4/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.4715 - accuracy: 0.4873\n",
      "Epoch 5/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.4713 - accuracy: 0.4870\n",
      "Epoch 6/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.4714 - accuracy: 0.4876\n",
      "Epoch 7/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.4716 - accuracy: 0.4877\n",
      "Epoch 8/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.4713 - accuracy: 0.4879\n",
      "Epoch 9/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.4715 - accuracy: 0.4881\n",
      "Epoch 10/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.4714 - accuracy: 0.4876\n",
      "Epoch 1/10\n",
      "150/150 [==============================] - 1s 2ms/step - loss: 0.5199 - accuracy: 0.4682\n",
      "Epoch 2/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.4959 - accuracy: 0.4836\n",
      "Epoch 3/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.4963 - accuracy: 0.4833\n",
      "Epoch 4/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.4959 - accuracy: 0.4843\n",
      "Epoch 5/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.4960 - accuracy: 0.4835\n",
      "Epoch 6/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.4959 - accuracy: 0.4838\n",
      "Epoch 7/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.4958 - accuracy: 0.4841\n",
      "Epoch 8/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.4959 - accuracy: 0.4838\n",
      "Epoch 9/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.4958 - accuracy: 0.4838\n",
      "Epoch 10/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.4958 - accuracy: 0.4840\n",
      "Epoch 1/10\n",
      "150/150 [==============================] - 1s 2ms/step - loss: 0.5450 - accuracy: 0.4658\n",
      "Epoch 2/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.5197 - accuracy: 0.4812\n",
      "Epoch 3/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.5199 - accuracy: 0.4812\n",
      "Epoch 4/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.5197 - accuracy: 0.4819\n",
      "Epoch 5/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.5199 - accuracy: 0.4815\n",
      "Epoch 6/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.5197 - accuracy: 0.4818\n",
      "Epoch 7/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.5197 - accuracy: 0.4816\n",
      "Epoch 8/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.5198 - accuracy: 0.4819\n",
      "Epoch 9/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.5198 - accuracy: 0.4813\n",
      "Epoch 10/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.5197 - accuracy: 0.4814\n",
      "Epoch 1/10\n",
      "150/150 [==============================] - 1s 2ms/step - loss: 0.5610 - accuracy: 0.4602\n",
      "Epoch 2/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.5427 - accuracy: 0.4758\n",
      "Epoch 3/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.5428 - accuracy: 0.4764\n",
      "Epoch 4/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.5428 - accuracy: 0.4764\n",
      "Epoch 5/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.5429 - accuracy: 0.4766\n",
      "Epoch 6/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.5427 - accuracy: 0.4771\n",
      "Epoch 7/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.5426 - accuracy: 0.4766\n",
      "Epoch 8/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.5428 - accuracy: 0.4776\n",
      "Epoch 9/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.5428 - accuracy: 0.4766\n",
      "Epoch 10/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.5426 - accuracy: 0.4761\n",
      "Epoch 1/10\n",
      "150/150 [==============================] - 1s 2ms/step - loss: 0.5789 - accuracy: 0.4602\n",
      "Epoch 2/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.5663 - accuracy: 0.4731\n",
      "Epoch 3/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.5662 - accuracy: 0.4737\n",
      "Epoch 4/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.5661 - accuracy: 0.4736\n",
      "Epoch 5/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.5661 - accuracy: 0.4744\n",
      "Epoch 6/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.5664 - accuracy: 0.4742\n",
      "Epoch 7/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.5662 - accuracy: 0.4738\n",
      "Epoch 8/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.5661 - accuracy: 0.4741\n",
      "Epoch 9/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.5660 - accuracy: 0.4742\n",
      "Epoch 10/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.5662 - accuracy: 0.4737\n",
      "Epoch 1/10\n",
      "150/150 [==============================] - 1s 2ms/step - loss: 0.5959 - accuracy: 0.4582\n",
      "Epoch 2/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.5860 - accuracy: 0.4707\n",
      "Epoch 3/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.5860 - accuracy: 0.4709\n",
      "Epoch 4/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.5860 - accuracy: 0.4717\n",
      "Epoch 5/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.5860 - accuracy: 0.4711\n",
      "Epoch 6/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.5860 - accuracy: 0.4719\n",
      "Epoch 7/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.5860 - accuracy: 0.4716\n",
      "Epoch 8/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.5860 - accuracy: 0.4718\n",
      "Epoch 9/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.5860 - accuracy: 0.4717\n",
      "Epoch 10/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.5860 - accuracy: 0.4713\n",
      "Epoch 1/10\n",
      "150/150 [==============================] - 1s 2ms/step - loss: 0.6148 - accuracy: 0.4525\n",
      "Epoch 2/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6071 - accuracy: 0.4654\n",
      "Epoch 3/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6071 - accuracy: 0.4659\n",
      "Epoch 4/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6071 - accuracy: 0.4655\n",
      "Epoch 5/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6070 - accuracy: 0.4656\n",
      "Epoch 6/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6071 - accuracy: 0.4648\n",
      "Epoch 7/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6070 - accuracy: 0.4670\n",
      "Epoch 8/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.6070 - accuracy: 0.4647\n",
      "Epoch 9/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.6071 - accuracy: 0.4658\n",
      "Epoch 10/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.6071 - accuracy: 0.4654\n",
      "Epoch 1/10\n",
      "150/150 [==============================] - 1s 2ms/step - loss: 0.6309 - accuracy: 0.4501\n",
      "Epoch 2/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6256 - accuracy: 0.4606\n",
      "Epoch 3/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6257 - accuracy: 0.4607\n",
      "Epoch 4/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6256 - accuracy: 0.4619\n",
      "Epoch 5/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6257 - accuracy: 0.4605\n",
      "Epoch 6/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6256 - accuracy: 0.4615\n",
      "Epoch 7/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6257 - accuracy: 0.4606\n",
      "Epoch 8/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6257 - accuracy: 0.4615\n",
      "Epoch 9/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6256 - accuracy: 0.4612\n",
      "Epoch 10/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6256 - accuracy: 0.4609\n",
      "Epoch 1/10\n",
      "150/150 [==============================] - 1s 2ms/step - loss: 0.6456 - accuracy: 0.4466\n",
      "Epoch 2/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6420 - accuracy: 0.4562\n",
      "Epoch 3/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6421 - accuracy: 0.4571\n",
      "Epoch 4/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.6420 - accuracy: 0.4561\n",
      "Epoch 5/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.6420 - accuracy: 0.4583\n",
      "Epoch 6/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.6420 - accuracy: 0.4561\n",
      "Epoch 7/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6420 - accuracy: 0.4568\n",
      "Epoch 8/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6420 - accuracy: 0.4561\n",
      "Epoch 9/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6420 - accuracy: 0.4559\n",
      "Epoch 10/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6419 - accuracy: 0.4573\n",
      "Epoch 1/10\n",
      "150/150 [==============================] - 1s 2ms/step - loss: 0.6571 - accuracy: 0.4485\n",
      "Epoch 2/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6561 - accuracy: 0.4507\n",
      "Epoch 3/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6561 - accuracy: 0.4517\n",
      "Epoch 4/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6561 - accuracy: 0.4508\n",
      "Epoch 5/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6561 - accuracy: 0.4527\n",
      "Epoch 6/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6561 - accuracy: 0.4517\n",
      "Epoch 7/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.6561 - accuracy: 0.4511\n",
      "Epoch 8/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.6561 - accuracy: 0.4505\n",
      "Epoch 9/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.6560 - accuracy: 0.4499\n",
      "Epoch 10/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.6561 - accuracy: 0.4497\n",
      "Epoch 1/10\n",
      "150/150 [==============================] - 1s 2ms/step - loss: 0.6698 - accuracy: 0.4440\n",
      "Epoch 2/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.6689 - accuracy: 0.4454\n",
      "Epoch 3/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.6689 - accuracy: 0.4458\n",
      "Epoch 4/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.6689 - accuracy: 0.4447\n",
      "Epoch 5/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6689 - accuracy: 0.4453\n",
      "Epoch 6/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.6689 - accuracy: 0.4429\n",
      "Epoch 7/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.6689 - accuracy: 0.4422\n",
      "Epoch 8/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.6689 - accuracy: 0.4439\n",
      "Epoch 9/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.6689 - accuracy: 0.4441\n",
      "Epoch 10/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.6689 - accuracy: 0.4430\n",
      "Epoch 1/10\n",
      "150/150 [==============================] - 1s 2ms/step - loss: 0.6792 - accuracy: 0.4361\n",
      "Epoch 2/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6788 - accuracy: 0.4394\n",
      "Epoch 3/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6788 - accuracy: 0.4387\n",
      "Epoch 4/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6787 - accuracy: 0.4396\n",
      "Epoch 5/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6787 - accuracy: 0.4416\n",
      "Epoch 6/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6788 - accuracy: 0.4387\n",
      "Epoch 7/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6788 - accuracy: 0.4388\n",
      "Epoch 8/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.6787 - accuracy: 0.4408\n",
      "Epoch 9/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.6787 - accuracy: 0.4387\n",
      "Epoch 10/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.6787 - accuracy: 0.4398\n",
      "Epoch 1/10\n",
      "150/150 [==============================] - 1s 2ms/step - loss: 0.6867 - accuracy: 0.4328\n",
      "Epoch 2/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6864 - accuracy: 0.4363\n",
      "Epoch 3/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6865 - accuracy: 0.4353\n",
      "Epoch 4/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6864 - accuracy: 0.4348\n",
      "Epoch 5/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6864 - accuracy: 0.4349\n",
      "Epoch 6/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.6864 - accuracy: 0.4363\n",
      "Epoch 7/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.6864 - accuracy: 0.4392\n",
      "Epoch 8/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.6864 - accuracy: 0.4351\n",
      "Epoch 9/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.6864 - accuracy: 0.4362\n",
      "Epoch 10/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.6864 - accuracy: 0.4365\n",
      "Epoch 1/10\n",
      "150/150 [==============================] - 1s 2ms/step - loss: 0.6910 - accuracy: 0.4382\n",
      "Epoch 2/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6908 - accuracy: 0.4320\n",
      "Epoch 3/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6907 - accuracy: 0.4280\n",
      "Epoch 4/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6908 - accuracy: 0.4335\n",
      "Epoch 5/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6907 - accuracy: 0.4330\n",
      "Epoch 6/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6908 - accuracy: 0.4354\n",
      "Epoch 7/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.6907 - accuracy: 0.4347\n",
      "Epoch 8/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.6907 - accuracy: 0.4343\n",
      "Epoch 9/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.6907 - accuracy: 0.4356\n",
      "Epoch 10/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.6907 - accuracy: 0.4376\n",
      "Epoch 1/10\n",
      "150/150 [==============================] - 1s 2ms/step - loss: 0.6936 - accuracy: 0.3967\n",
      "Epoch 2/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6935 - accuracy: 0.3669\n",
      "Epoch 3/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6935 - accuracy: 0.3924\n",
      "Epoch 4/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6935 - accuracy: 0.3885\n",
      "Epoch 5/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6935 - accuracy: 0.3738\n",
      "Epoch 6/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6935 - accuracy: 0.3718\n",
      "Epoch 7/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6934 - accuracy: 0.4184\n",
      "Epoch 8/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6935 - accuracy: 0.3570\n",
      "Epoch 9/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6935 - accuracy: 0.3732\n",
      "Epoch 10/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6934 - accuracy: 0.3678\n",
      "Epoch 1/10\n",
      "150/150 [==============================] - 1s 2ms/step - loss: 0.6929 - accuracy: 0.3278\n",
      "Epoch 2/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6928 - accuracy: 0.3339\n",
      "Epoch 3/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6928 - accuracy: 0.3354\n",
      "Epoch 4/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6928 - accuracy: 0.3252\n",
      "Epoch 5/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6927 - accuracy: 0.3340\n",
      "Epoch 6/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.6927 - accuracy: 0.3395\n",
      "Epoch 7/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.6927 - accuracy: 0.3370\n",
      "Epoch 8/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.6927 - accuracy: 0.3366\n",
      "Epoch 9/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.6927 - accuracy: 0.3335\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 10/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.6927 - accuracy: 0.3386\n",
      "Epoch 1/10\n",
      "150/150 [==============================] - 1s 2ms/step - loss: 0.6893 - accuracy: 0.3379\n",
      "Epoch 2/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6889 - accuracy: 0.3340\n",
      "Epoch 3/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6889 - accuracy: 0.3337\n",
      "Epoch 4/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6889 - accuracy: 0.3335\n",
      "Epoch 5/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6889 - accuracy: 0.3317\n",
      "Epoch 6/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.6889 - accuracy: 0.3380\n",
      "Epoch 7/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.6889 - accuracy: 0.3347\n",
      "Epoch 8/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.6889 - accuracy: 0.3348\n",
      "Epoch 9/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6889 - accuracy: 0.3381\n",
      "Epoch 10/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6889 - accuracy: 0.3310\n",
      "Epoch 1/10\n",
      "150/150 [==============================] - 1s 2ms/step - loss: 0.6838 - accuracy: 0.3378\n",
      "Epoch 2/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6837 - accuracy: 0.3446\n",
      "Epoch 3/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6837 - accuracy: 0.3426\n",
      "Epoch 4/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6837 - accuracy: 0.3405\n",
      "Epoch 5/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6836 - accuracy: 0.3418\n",
      "Epoch 6/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6836 - accuracy: 0.3406\n",
      "Epoch 7/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6837 - accuracy: 0.3396\n",
      "Epoch 8/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6837 - accuracy: 0.3392\n",
      "Epoch 9/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6837 - accuracy: 0.3415\n",
      "Epoch 10/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6837 - accuracy: 0.3427\n",
      "Epoch 1/10\n",
      "150/150 [==============================] - 1s 2ms/step - loss: 0.6757 - accuracy: 0.3418\n",
      "Epoch 2/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6745 - accuracy: 0.3476\n",
      "Epoch 3/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6744 - accuracy: 0.3445\n",
      "Epoch 4/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6744 - accuracy: 0.3432\n",
      "Epoch 5/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6744 - accuracy: 0.3467\n",
      "Epoch 6/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6744 - accuracy: 0.3446\n",
      "Epoch 7/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.6744 - accuracy: 0.3439\n",
      "Epoch 8/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.6743 - accuracy: 0.3457\n",
      "Epoch 9/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.6744 - accuracy: 0.3437\n",
      "Epoch 10/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.6744 - accuracy: 0.3447\n",
      "Epoch 1/10\n",
      "150/150 [==============================] - 1s 2ms/step - loss: 0.6655 - accuracy: 0.3475\n",
      "Epoch 2/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6633 - accuracy: 0.3522\n",
      "Epoch 3/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6635 - accuracy: 0.3553\n",
      "Epoch 4/10\n",
      "150/150 [==============================] - 0s 2ms/step - loss: 0.6635 - accuracy: 0.3551\n",
      "Epoch 5/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.6633 - accuracy: 0.3520\n",
      "Epoch 6/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.6634 - accuracy: 0.3536\n",
      "Epoch 7/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.6634 - accuracy: 0.3538\n",
      "Epoch 8/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.6634 - accuracy: 0.3533\n",
      "Epoch 9/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.6634 - accuracy: 0.3529\n",
      "Epoch 10/10\n",
      "150/150 [==============================] - 0s 3ms/step - loss: 0.6634 - accuracy: 0.3513\n"
     ]
    }
   ],
   "source": [
    "losses = []\n",
    "for lambda1 in np.linspace(-1,1,20):\n",
    "\n",
    "    xvals_1 = np.concatenate([gauss_data,gauss_sim])\n",
    "    yvals_1 = np.concatenate([np.ones(len(gauss_data)),np.zeros(len(gauss_sim))])\n",
    "    weights_1 = np.concatenate([np.ones(len(gauss_data)),np.exp(lambda1*gauss_sim)*len(gauss_data)/np.sum(np.exp(lambda1*gauss_sim))])\n",
    "\n",
    "    X_train_1, X_test_1, Y_train_1, Y_test_1, w_train_1, w_test_1 = train_test_split(xvals_1, yvals_1, weights_1)\n",
    "\n",
    "    Y_train_2 = np.stack((Y_train_1, w_train_1), axis=1)\n",
    "    Y_test_2 = np.stack((Y_test_1, w_test_1), axis=1)\n",
    "\n",
    "    inputs = Input((1, ))\n",
    "    hidden_layer_1 = Dense(50, activation='relu')(inputs)\n",
    "    hidden_layer_2 = Dense(50, activation='relu')(hidden_layer_1)\n",
    "    hidden_layer_3 = Dense(50, activation='relu')(hidden_layer_2)\n",
    "    outputs = Dense(1, activation='sigmoid')(hidden_layer_3)\n",
    "    model = Model(inputs=inputs, outputs=outputs)\n",
    "    \n",
    "    model.compile(loss=weighted_binary_crossentropy, optimizer='Adam', metrics=['accuracy'])\n",
    "    model.fit(X_train_1,\n",
    "              Y_train_2,\n",
    "              epochs=10,\n",
    "              batch_size=1000,\n",
    "              verbose=1)\n",
    "    losses+=[model.history.history['loss'][-1]]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "862e8935",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0.0, 8000.0)"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "lambda1 = 1\n",
    "\n",
    "xvals_1 = np.concatenate([gauss_data,gauss_sim])\n",
    "yvals_1 = np.concatenate([np.ones(len(gauss_data)),np.zeros(len(gauss_sim))])\n",
    "weights_1 = np.concatenate([np.ones(len(gauss_data)),np.exp(lambda1*gauss_sim)*len(gauss_data)/np.sum(np.exp(lambda1*gauss_sim))])\n",
    "\n",
    "X_train_1, X_test_1, Y_train_1, Y_test_1, w_train_1, w_test_1 = train_test_split(xvals_1, yvals_1, weights_1)\n",
    "\n",
    "_,_,_=plt.hist(X_test_1[Y_test_1==1],bins=np.linspace(-4,4,20),alpha=0.5,label=\"data\")\n",
    "_,_,_=plt.hist(X_test_1[Y_test_1==0],bins=np.linspace(-4,4,20),alpha=0.5,label=\"MC\")\n",
    "_,_,_=plt.hist(X_test_1[Y_test_1==0],bins=np.linspace(-4,4,20),weights=w_test_1[Y_test_1==0],histtype=\"step\",color=\"black\",ls=\":\",label=\"weighted MC\")\n",
    "plt.legend(fontsize=15)\n",
    "plt.ylim([0,8000])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "0aeb2d3e",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x7fc62a6b57c0>]"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(np.linspace(-1,1,20),losses)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "784e9bae",
   "metadata": {},
   "outputs": [],
   "source": [
    "#GW,_,_ = plt.hist(w_true,weights=weightsW[-1,1],bins=np.linspace(0,1,30),histtype=\"step\",ls=\"-\",color=\"black\",label=\"rw\")\n",
    "#OFW,_,_  = plt.hist(w_true_alt,bins=np.linspace(0,1,30),ls=\":\",histtype=\"step\",color=\"black\",lw=3,label=\"target\")\n",
    "#TW,_,_  = plt.hist(w_true,bins=np.linspace(0,1,30),alpha=0.2,color=\"blue\",label=\"prior\")\n",
    "#plt.legend()\n",
    "#plt.xlabel(\"jet width\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "99aa6332",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/10\n",
      "1500/1500 [==============================] - 4s 2ms/step - loss: 0.6672 - accuracy: 0.4193\n",
      "Epoch 2/10\n",
      "1500/1500 [==============================] - 4s 2ms/step - loss: 0.6663 - accuracy: 0.4193\n",
      "Epoch 3/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6663 - accuracy: 0.4183\n",
      "Epoch 4/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6662 - accuracy: 0.4183\n",
      "Epoch 5/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6662 - accuracy: 0.4184\n",
      "Epoch 6/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6662 - accuracy: 0.4182\n",
      "Epoch 7/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6662 - accuracy: 0.4179\n",
      "Epoch 8/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6662 - accuracy: 0.4183\n",
      "Epoch 9/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6662 - accuracy: 0.4179\n",
      "Epoch 10/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6662 - accuracy: 0.4181\n",
      "Epoch 1/10\n",
      "1500/1500 [==============================] - 4s 2ms/step - loss: 0.6725 - accuracy: 0.4181\n",
      "Epoch 2/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6720 - accuracy: 0.4192\n",
      "Epoch 3/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6720 - accuracy: 0.4190\n",
      "Epoch 4/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6719 - accuracy: 0.4188\n",
      "Epoch 5/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6719 - accuracy: 0.4185\n",
      "Epoch 6/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6719 - accuracy: 0.4183\n",
      "Epoch 7/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6719 - accuracy: 0.4184\n",
      "Epoch 8/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6719 - accuracy: 0.4181\n",
      "Epoch 9/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6719 - accuracy: 0.4185\n",
      "Epoch 10/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6719 - accuracy: 0.4185\n",
      "Epoch 1/10\n",
      "1500/1500 [==============================] - 3s 2ms/step - loss: 0.6775 - accuracy: 0.4191\n",
      "Epoch 2/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6769 - accuracy: 0.4212\n",
      "Epoch 3/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6769 - accuracy: 0.4210\n",
      "Epoch 4/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6768 - accuracy: 0.4211\n",
      "Epoch 5/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6768 - accuracy: 0.4207\n",
      "Epoch 6/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6768 - accuracy: 0.4204\n",
      "Epoch 7/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6768 - accuracy: 0.4209\n",
      "Epoch 8/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6768 - accuracy: 0.4205\n",
      "Epoch 9/10\n",
      "1500/1500 [==============================] - 4s 2ms/step - loss: 0.6768 - accuracy: 0.4201\n",
      "Epoch 10/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6768 - accuracy: 0.4200\n",
      "Epoch 1/10\n",
      "1500/1500 [==============================] - 4s 2ms/step - loss: 0.6819 - accuracy: 0.4193\n",
      "Epoch 2/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6814 - accuracy: 0.4222\n",
      "Epoch 3/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6814 - accuracy: 0.4223\n",
      "Epoch 4/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6814 - accuracy: 0.4217\n",
      "Epoch 5/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6814 - accuracy: 0.4212\n",
      "Epoch 6/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6814 - accuracy: 0.4216\n",
      "Epoch 7/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6814 - accuracy: 0.4217\n",
      "Epoch 8/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6814 - accuracy: 0.4220\n",
      "Epoch 9/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6814 - accuracy: 0.4217\n",
      "Epoch 10/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6814 - accuracy: 0.4216\n",
      "Epoch 1/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6856 - accuracy: 0.4249\n",
      "Epoch 2/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6853 - accuracy: 0.4268\n",
      "Epoch 3/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6852 - accuracy: 0.4265\n",
      "Epoch 4/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6852 - accuracy: 0.4251\n",
      "Epoch 5/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6852 - accuracy: 0.4262\n",
      "Epoch 6/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6852 - accuracy: 0.4249\n",
      "Epoch 7/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6852 - accuracy: 0.4263\n",
      "Epoch 8/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6852 - accuracy: 0.4260\n",
      "Epoch 9/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6852 - accuracy: 0.4257\n",
      "Epoch 10/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6852 - accuracy: 0.4255\n",
      "Epoch 1/10\n",
      "1500/1500 [==============================] - 3s 2ms/step - loss: 0.6885 - accuracy: 0.4283\n",
      "Epoch 2/10\n",
      "1500/1500 [==============================] - 3s 2ms/step - loss: 0.6882 - accuracy: 0.4324\n",
      "Epoch 3/10\n",
      "1500/1500 [==============================] - 3s 2ms/step - loss: 0.6882 - accuracy: 0.4311\n",
      "Epoch 4/10\n",
      "1500/1500 [==============================] - 3s 2ms/step - loss: 0.6882 - accuracy: 0.4306\n",
      "Epoch 5/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6882 - accuracy: 0.4308\n",
      "Epoch 6/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6882 - accuracy: 0.4311\n",
      "Epoch 7/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6881 - accuracy: 0.4306\n",
      "Epoch 8/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6881 - accuracy: 0.4313\n",
      "Epoch 9/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6881 - accuracy: 0.4306\n",
      "Epoch 10/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6881 - accuracy: 0.4303\n",
      "Epoch 1/10\n",
      "1500/1500 [==============================] - 3s 2ms/step - loss: 0.6905 - accuracy: 0.4231\n",
      "Epoch 2/10\n",
      "1500/1500 [==============================] - 4s 2ms/step - loss: 0.6902 - accuracy: 0.4242\n",
      "Epoch 3/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6902 - accuracy: 0.4254\n",
      "Epoch 4/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6901 - accuracy: 0.4239\n",
      "Epoch 5/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6901 - accuracy: 0.4235\n",
      "Epoch 6/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6901 - accuracy: 0.4246\n",
      "Epoch 7/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6901 - accuracy: 0.4243\n",
      "Epoch 8/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6901 - accuracy: 0.4236\n",
      "Epoch 9/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6901 - accuracy: 0.4240\n",
      "Epoch 10/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6901 - accuracy: 0.4238\n",
      "Epoch 1/10\n",
      "1500/1500 [==============================] - 4s 2ms/step - loss: 0.6910 - accuracy: 0.4046\n",
      "Epoch 2/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6907 - accuracy: 0.4188\n",
      "Epoch 3/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6907 - accuracy: 0.4198\n",
      "Epoch 4/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6907 - accuracy: 0.4217\n",
      "Epoch 5/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6907 - accuracy: 0.4208\n",
      "Epoch 6/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6907 - accuracy: 0.4211\n",
      "Epoch 7/10\n",
      "1500/1500 [==============================] - 4s 2ms/step - loss: 0.6907 - accuracy: 0.4218\n",
      "Epoch 8/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6907 - accuracy: 0.4227\n",
      "Epoch 9/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6907 - accuracy: 0.4207\n",
      "Epoch 10/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6906 - accuracy: 0.4227\n",
      "Epoch 1/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6903 - accuracy: 0.4142\n",
      "Epoch 2/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6901 - accuracy: 0.4144\n",
      "Epoch 3/10\n",
      "1500/1500 [==============================] - 4s 2ms/step - loss: 0.6901 - accuracy: 0.4149\n",
      "Epoch 4/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6901 - accuracy: 0.4168\n",
      "Epoch 5/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6901 - accuracy: 0.4161\n",
      "Epoch 6/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6901 - accuracy: 0.4163\n",
      "Epoch 7/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6901 - accuracy: 0.4163\n",
      "Epoch 8/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6901 - accuracy: 0.4165\n",
      "Epoch 9/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6901 - accuracy: 0.4155\n",
      "Epoch 10/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6900 - accuracy: 0.4163\n",
      "Epoch 1/10\n",
      "1500/1500 [==============================] - 4s 2ms/step - loss: 0.6884 - accuracy: 0.4253\n",
      "Epoch 2/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6880 - accuracy: 0.4205\n",
      "Epoch 3/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6880 - accuracy: 0.4193\n",
      "Epoch 4/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6880 - accuracy: 0.4194\n",
      "Epoch 5/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6879 - accuracy: 0.4202\n",
      "Epoch 6/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6879 - accuracy: 0.4194\n",
      "Epoch 7/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6879 - accuracy: 0.4193\n",
      "Epoch 8/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6879 - accuracy: 0.4188\n",
      "Epoch 9/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6879 - accuracy: 0.4186\n",
      "Epoch 10/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6879 - accuracy: 0.4192\n",
      "Epoch 1/10\n",
      "1500/1500 [==============================] - 3s 2ms/step - loss: 0.6845 - accuracy: 0.4297\n",
      "Epoch 2/10\n",
      "1500/1500 [==============================] - 3s 2ms/step - loss: 0.6840 - accuracy: 0.4341\n",
      "Epoch 3/10\n",
      "1500/1500 [==============================] - 3s 2ms/step - loss: 0.6840 - accuracy: 0.4342\n",
      "Epoch 4/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6839 - accuracy: 0.4344\n",
      "Epoch 5/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6840 - accuracy: 0.4343\n",
      "Epoch 6/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6840 - accuracy: 0.4348\n",
      "Epoch 7/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6839 - accuracy: 0.4336\n",
      "Epoch 8/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6839 - accuracy: 0.4356\n",
      "Epoch 9/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6839 - accuracy: 0.4341\n",
      "Epoch 10/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6839 - accuracy: 0.4344\n",
      "Epoch 1/10\n",
      "1500/1500 [==============================] - 4s 2ms/step - loss: 0.6786 - accuracy: 0.4271\n",
      "Epoch 2/10\n",
      "1500/1500 [==============================] - 3s 2ms/step - loss: 0.6782 - accuracy: 0.4299\n",
      "Epoch 3/10\n",
      "1500/1500 [==============================] - 3s 2ms/step - loss: 0.6781 - accuracy: 0.4293\n",
      "Epoch 4/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6781 - accuracy: 0.4287\n",
      "Epoch 5/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6781 - accuracy: 0.4294\n",
      "Epoch 6/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6781 - accuracy: 0.4283\n",
      "Epoch 7/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6781 - accuracy: 0.4290\n",
      "Epoch 8/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6781 - accuracy: 0.4290\n",
      "Epoch 9/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6781 - accuracy: 0.4291\n",
      "Epoch 10/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6780 - accuracy: 0.4291\n",
      "Epoch 1/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6711 - accuracy: 0.4270\n",
      "Epoch 2/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6704 - accuracy: 0.4273\n",
      "Epoch 3/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6704 - accuracy: 0.4276\n",
      "Epoch 4/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6704 - accuracy: 0.4273\n",
      "Epoch 5/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6704 - accuracy: 0.4265\n",
      "Epoch 6/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6704 - accuracy: 0.4280\n",
      "Epoch 7/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6704 - accuracy: 0.4268\n",
      "Epoch 8/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6704 - accuracy: 0.4275\n",
      "Epoch 9/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6704 - accuracy: 0.4273\n",
      "Epoch 10/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6704 - accuracy: 0.4280\n",
      "Epoch 1/10\n",
      "1500/1500 [==============================] - 4s 2ms/step - loss: 0.6610 - accuracy: 0.4221\n",
      "Epoch 2/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6603 - accuracy: 0.4263\n",
      "Epoch 3/10\n",
      "1500/1500 [==============================] - 4s 2ms/step - loss: 0.6603 - accuracy: 0.4251\n",
      "Epoch 4/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6603 - accuracy: 0.4252\n",
      "Epoch 5/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6603 - accuracy: 0.4251\n",
      "Epoch 6/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6603 - accuracy: 0.4253\n",
      "Epoch 7/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6603 - accuracy: 0.4250\n",
      "Epoch 8/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6603 - accuracy: 0.4257\n",
      "Epoch 9/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6602 - accuracy: 0.4254\n",
      "Epoch 10/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6603 - accuracy: 0.4252\n",
      "Epoch 1/10\n",
      "1500/1500 [==============================] - 3s 2ms/step - loss: 0.6496 - accuracy: 0.4231\n",
      "Epoch 2/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6484 - accuracy: 0.4272\n",
      "Epoch 3/10\n",
      "1500/1500 [==============================] - 4s 2ms/step - loss: 0.6483 - accuracy: 0.4267\n",
      "Epoch 4/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6483 - accuracy: 0.4269\n",
      "Epoch 5/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6484 - accuracy: 0.4267\n",
      "Epoch 6/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6483 - accuracy: 0.4267\n",
      "Epoch 7/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6483 - accuracy: 0.4265\n",
      "Epoch 8/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6483 - accuracy: 0.4271\n",
      "Epoch 9/10\n",
      "1500/1500 [==============================] - 4s 2ms/step - loss: 0.6483 - accuracy: 0.4268\n",
      "Epoch 10/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6483 - accuracy: 0.4267\n",
      "Epoch 1/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6352 - accuracy: 0.4252\n",
      "Epoch 2/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6340 - accuracy: 0.4272\n",
      "Epoch 3/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6340 - accuracy: 0.4267\n",
      "Epoch 4/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6339 - accuracy: 0.4265\n",
      "Epoch 5/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6339 - accuracy: 0.4266\n",
      "Epoch 6/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6339 - accuracy: 0.4264\n",
      "Epoch 7/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6339 - accuracy: 0.4265\n",
      "Epoch 8/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6339 - accuracy: 0.4261\n",
      "Epoch 9/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6339 - accuracy: 0.4265\n",
      "Epoch 10/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6339 - accuracy: 0.4265\n",
      "Epoch 1/10\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6198 - accuracy: 0.4237\n",
      "Epoch 2/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6179 - accuracy: 0.4288\n",
      "Epoch 3/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6178 - accuracy: 0.4286\n",
      "Epoch 4/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6178 - accuracy: 0.4283\n",
      "Epoch 5/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6178 - accuracy: 0.4283\n",
      "Epoch 6/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6178 - accuracy: 0.4284\n",
      "Epoch 7/10\n",
      "1500/1500 [==============================] - 3s 2ms/step - loss: 0.6178 - accuracy: 0.4282\n",
      "Epoch 8/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6177 - accuracy: 0.4288\n",
      "Epoch 9/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6178 - accuracy: 0.4287\n",
      "Epoch 10/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6177 - accuracy: 0.4286\n",
      "Epoch 1/10\n",
      "1500/1500 [==============================] - 4s 2ms/step - loss: 0.6023 - accuracy: 0.4248\n",
      "Epoch 2/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.5995 - accuracy: 0.4303\n",
      "Epoch 3/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.5994 - accuracy: 0.4299\n",
      "Epoch 4/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.5995 - accuracy: 0.4301\n",
      "Epoch 5/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.5994 - accuracy: 0.4300\n",
      "Epoch 6/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.5994 - accuracy: 0.4301\n",
      "Epoch 7/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.5994 - accuracy: 0.4299\n",
      "Epoch 8/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.5994 - accuracy: 0.4298\n",
      "Epoch 9/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.5994 - accuracy: 0.4303\n",
      "Epoch 10/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.5994 - accuracy: 0.4302\n",
      "Epoch 1/10\n",
      "1500/1500 [==============================] - 4s 2ms/step - loss: 0.5828 - accuracy: 0.4286\n",
      "Epoch 2/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.5794 - accuracy: 0.4331\n",
      "Epoch 3/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.5793 - accuracy: 0.4331\n",
      "Epoch 4/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.5793 - accuracy: 0.4331\n",
      "Epoch 5/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.5793 - accuracy: 0.4330\n",
      "Epoch 6/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.5793 - accuracy: 0.4329\n",
      "Epoch 7/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.5793 - accuracy: 0.4331\n",
      "Epoch 8/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.5793 - accuracy: 0.4332\n",
      "Epoch 9/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.5793 - accuracy: 0.4333\n",
      "Epoch 10/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.5793 - accuracy: 0.4330\n",
      "Epoch 1/10\n",
      "1500/1500 [==============================] - 4s 2ms/step - loss: 0.5628 - accuracy: 0.4296\n",
      "Epoch 2/10\n",
      "1500/1500 [==============================] - 4s 2ms/step - loss: 0.5578 - accuracy: 0.4353\n",
      "Epoch 3/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.5578 - accuracy: 0.4354\n",
      "Epoch 4/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.5577 - accuracy: 0.4354\n",
      "Epoch 5/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.5577 - accuracy: 0.4354\n",
      "Epoch 6/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.5577 - accuracy: 0.4356\n",
      "Epoch 7/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.5577 - accuracy: 0.4355\n",
      "Epoch 8/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.5577 - accuracy: 0.4354\n",
      "Epoch 9/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.5577 - accuracy: 0.4359\n",
      "Epoch 10/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.5577 - accuracy: 0.4355\n"
     ]
    }
   ],
   "source": [
    "losses_width = []\n",
    "for lambda1 in np.linspace(-2,10,20):\n",
    "\n",
    "    xvals_1 = np.concatenate([w_true_alt,w_true])\n",
    "    yvals_1 = np.concatenate([np.ones(len(w_true_alt)),np.zeros(len(w_true))])\n",
    "    weights_1 = np.concatenate([np.ones(len(w_true_alt)),np.exp(lambda1*w_true)*len(w_true_alt)/np.sum(np.exp(lambda1*w_true))])\n",
    "\n",
    "    X_train_1, X_test_1, Y_train_1, Y_test_1, w_train_1, w_test_1 = train_test_split(xvals_1, yvals_1, weights_1)\n",
    "\n",
    "    Y_train_2 = np.stack((Y_train_1, w_train_1), axis=1)\n",
    "    Y_test_2 = np.stack((Y_test_1, w_test_1), axis=1)\n",
    "\n",
    "    inputs = Input((1, ))\n",
    "    hidden_layer_1 = Dense(50, activation='relu')(inputs)\n",
    "    hidden_layer_2 = Dense(50, activation='relu')(hidden_layer_1)\n",
    "    hidden_layer_3 = Dense(50, activation='relu')(hidden_layer_2)\n",
    "    outputs = Dense(1, activation='sigmoid')(hidden_layer_3)\n",
    "    model = Model(inputs=inputs, outputs=outputs)\n",
    "    \n",
    "    model.compile(loss=weighted_binary_crossentropy, optimizer='Adam', metrics=['accuracy'])\n",
    "    model.fit(X_train_1,\n",
    "              Y_train_2,\n",
    "              epochs=10,\n",
    "              batch_size=1000,\n",
    "              verbose=1)\n",
    "    losses_width+=[model.history.history['loss'][-1]]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "id": "ea2460bb",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/10\n",
      "1500/1500 [==============================] - 4s 2ms/step - loss: 0.6777 - accuracy: 0.4159\n",
      "Epoch 2/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6773 - accuracy: 0.4182\n",
      "Epoch 3/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6773 - accuracy: 0.4171\n",
      "Epoch 4/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6773 - accuracy: 0.4170\n",
      "Epoch 5/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6772 - accuracy: 0.4169\n",
      "Epoch 6/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6772 - accuracy: 0.4168\n",
      "Epoch 7/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6772 - accuracy: 0.4166\n",
      "Epoch 8/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6772 - accuracy: 0.4171\n",
      "Epoch 9/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6772 - accuracy: 0.4170\n",
      "Epoch 10/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6772 - accuracy: 0.4168\n",
      "Epoch 1/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6792 - accuracy: 0.4184\n",
      "Epoch 2/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6787 - accuracy: 0.4197\n",
      "Epoch 3/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6787 - accuracy: 0.4182\n",
      "Epoch 4/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6787 - accuracy: 0.4184\n",
      "Epoch 5/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6787 - accuracy: 0.4183\n",
      "Epoch 6/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6787 - accuracy: 0.4186\n",
      "Epoch 7/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6787 - accuracy: 0.4185\n",
      "Epoch 8/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6787 - accuracy: 0.4183\n",
      "Epoch 9/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6787 - accuracy: 0.4183\n",
      "Epoch 10/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6786 - accuracy: 0.4177\n",
      "Epoch 1/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6808 - accuracy: 0.4200\n",
      "Epoch 2/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6805 - accuracy: 0.4207\n",
      "Epoch 3/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6805 - accuracy: 0.4211\n",
      "Epoch 4/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6805 - accuracy: 0.4198\n",
      "Epoch 5/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6804 - accuracy: 0.4199\n",
      "Epoch 6/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6804 - accuracy: 0.4209\n",
      "Epoch 7/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6804 - accuracy: 0.4202\n",
      "Epoch 8/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6804 - accuracy: 0.4201\n",
      "Epoch 9/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6804 - accuracy: 0.4203\n",
      "Epoch 10/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6804 - accuracy: 0.4203\n",
      "Epoch 1/10\n",
      "1500/1500 [==============================] - 4s 2ms/step - loss: 0.6823 - accuracy: 0.4199\n",
      "Epoch 2/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6820 - accuracy: 0.4210\n",
      "Epoch 3/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6819 - accuracy: 0.4212\n",
      "Epoch 4/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6819 - accuracy: 0.4214\n",
      "Epoch 5/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6819 - accuracy: 0.4215\n",
      "Epoch 6/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6819 - accuracy: 0.4210\n",
      "Epoch 7/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6819 - accuracy: 0.4209\n",
      "Epoch 8/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6819 - accuracy: 0.4210\n",
      "Epoch 9/10\n",
      "1500/1500 [==============================] - 4s 2ms/step - loss: 0.6819 - accuracy: 0.4208\n",
      "Epoch 10/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6819 - accuracy: 0.4212\n",
      "Epoch 1/10\n",
      "1500/1500 [==============================] - 4s 2ms/step - loss: 0.6837 - accuracy: 0.4229\n",
      "Epoch 2/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6833 - accuracy: 0.4261\n",
      "Epoch 3/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6832 - accuracy: 0.4253\n",
      "Epoch 4/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6832 - accuracy: 0.4257\n",
      "Epoch 5/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6832 - accuracy: 0.4248\n",
      "Epoch 6/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6832 - accuracy: 0.4248\n",
      "Epoch 7/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6832 - accuracy: 0.4248\n",
      "Epoch 8/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6832 - accuracy: 0.4249\n",
      "Epoch 9/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6832 - accuracy: 0.4250\n",
      "Epoch 10/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6832 - accuracy: 0.4249\n",
      "Epoch 1/10\n",
      "1500/1500 [==============================] - 3s 2ms/step - loss: 0.6849 - accuracy: 0.4244\n",
      "Epoch 2/10\n",
      "1500/1500 [==============================] - 3s 2ms/step - loss: 0.6846 - accuracy: 0.4268\n",
      "Epoch 3/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6846 - accuracy: 0.4265\n",
      "Epoch 4/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6845 - accuracy: 0.4258\n",
      "Epoch 5/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6845 - accuracy: 0.4251\n",
      "Epoch 6/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6845 - accuracy: 0.4258\n",
      "Epoch 7/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6845 - accuracy: 0.4262\n",
      "Epoch 8/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6845 - accuracy: 0.4268\n",
      "Epoch 9/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6845 - accuracy: 0.4268\n",
      "Epoch 10/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6845 - accuracy: 0.4260\n",
      "Epoch 1/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6861 - accuracy: 0.4251\n",
      "Epoch 2/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6857 - accuracy: 0.4294\n",
      "Epoch 3/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6856 - accuracy: 0.4293\n",
      "Epoch 4/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6856 - accuracy: 0.4282\n",
      "Epoch 5/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6856 - accuracy: 0.4286\n",
      "Epoch 6/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6856 - accuracy: 0.4288\n",
      "Epoch 7/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6856 - accuracy: 0.4286\n",
      "Epoch 8/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6856 - accuracy: 0.4285\n",
      "Epoch 9/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6856 - accuracy: 0.4287\n",
      "Epoch 10/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6856 - accuracy: 0.4290\n",
      "Epoch 1/10\n",
      "1500/1500 [==============================] - 4s 2ms/step - loss: 0.6869 - accuracy: 0.4302\n",
      "Epoch 2/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6865 - accuracy: 0.4307\n",
      "Epoch 3/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6864 - accuracy: 0.4292\n",
      "Epoch 4/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6864 - accuracy: 0.4298\n",
      "Epoch 5/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6864 - accuracy: 0.4300\n",
      "Epoch 6/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6864 - accuracy: 0.4301\n",
      "Epoch 7/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6864 - accuracy: 0.4294\n",
      "Epoch 8/10\n",
      "1500/1500 [==============================] - 4s 2ms/step - loss: 0.6864 - accuracy: 0.4295\n",
      "Epoch 9/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6864 - accuracy: 0.4290\n",
      "Epoch 10/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6864 - accuracy: 0.4286\n",
      "Epoch 1/10\n",
      "1500/1500 [==============================] - 4s 2ms/step - loss: 0.6877 - accuracy: 0.4238\n",
      "Epoch 2/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6871 - accuracy: 0.4257\n",
      "Epoch 3/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6871 - accuracy: 0.4268\n",
      "Epoch 4/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6871 - accuracy: 0.4274\n",
      "Epoch 5/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6871 - accuracy: 0.4266\n",
      "Epoch 6/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6871 - accuracy: 0.4267\n",
      "Epoch 7/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6870 - accuracy: 0.4271\n",
      "Epoch 8/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6870 - accuracy: 0.4267\n",
      "Epoch 9/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6870 - accuracy: 0.4265\n",
      "Epoch 10/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6870 - accuracy: 0.4264\n",
      "Epoch 1/10\n",
      "1500/1500 [==============================] - 3s 2ms/step - loss: 0.6881 - accuracy: 0.4191\n",
      "Epoch 2/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6877 - accuracy: 0.4231\n",
      "Epoch 3/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6876 - accuracy: 0.4228\n",
      "Epoch 4/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6876 - accuracy: 0.4250\n",
      "Epoch 5/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6876 - accuracy: 0.4254\n",
      "Epoch 6/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6876 - accuracy: 0.4256\n",
      "Epoch 7/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6876 - accuracy: 0.4244\n",
      "Epoch 8/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6876 - accuracy: 0.4249\n",
      "Epoch 9/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6876 - accuracy: 0.4255\n",
      "Epoch 10/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6876 - accuracy: 0.4253\n",
      "Epoch 1/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6882 - accuracy: 0.4203\n",
      "Epoch 2/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6878 - accuracy: 0.4214\n",
      "Epoch 3/10\n",
      "1500/1500 [==============================] - 4s 2ms/step - loss: 0.6878 - accuracy: 0.4221\n",
      "Epoch 4/10\n",
      "1500/1500 [==============================] - 4s 2ms/step - loss: 0.6878 - accuracy: 0.4219\n",
      "Epoch 5/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6878 - accuracy: 0.4220\n",
      "Epoch 6/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6878 - accuracy: 0.4219\n",
      "Epoch 7/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6878 - accuracy: 0.4215\n",
      "Epoch 8/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6878 - accuracy: 0.4227\n",
      "Epoch 9/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6878 - accuracy: 0.4227\n",
      "Epoch 10/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6878 - accuracy: 0.4230\n",
      "Epoch 1/10\n",
      "1500/1500 [==============================] - 4s 2ms/step - loss: 0.6881 - accuracy: 0.4171\n",
      "Epoch 2/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6876 - accuracy: 0.4252\n",
      "Epoch 3/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6876 - accuracy: 0.4264\n",
      "Epoch 4/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6876 - accuracy: 0.4256\n",
      "Epoch 5/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6876 - accuracy: 0.4264\n",
      "Epoch 6/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6876 - accuracy: 0.4257\n",
      "Epoch 7/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6876 - accuracy: 0.4265\n",
      "Epoch 8/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6876 - accuracy: 0.4259\n",
      "Epoch 9/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6876 - accuracy: 0.4257\n",
      "Epoch 10/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6876 - accuracy: 0.4272\n",
      "Epoch 1/10\n",
      "1500/1500 [==============================] - 4s 2ms/step - loss: 0.6878 - accuracy: 0.4152\n",
      "Epoch 2/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6873 - accuracy: 0.4228\n",
      "Epoch 3/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6873 - accuracy: 0.4242\n",
      "Epoch 4/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6872 - accuracy: 0.4250\n",
      "Epoch 5/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6872 - accuracy: 0.4253\n",
      "Epoch 6/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6872 - accuracy: 0.4258\n",
      "Epoch 7/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6872 - accuracy: 0.4257\n",
      "Epoch 8/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6872 - accuracy: 0.4257\n",
      "Epoch 9/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6872 - accuracy: 0.4256\n",
      "Epoch 10/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6872 - accuracy: 0.4263\n",
      "Epoch 1/10\n",
      "1500/1500 [==============================] - 4s 2ms/step - loss: 0.6869 - accuracy: 0.4177\n",
      "Epoch 2/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6863 - accuracy: 0.4237\n",
      "Epoch 3/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6863 - accuracy: 0.4239\n",
      "Epoch 4/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6863 - accuracy: 0.4255\n",
      "Epoch 5/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6863 - accuracy: 0.4249\n",
      "Epoch 6/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6862 - accuracy: 0.4256\n",
      "Epoch 7/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6863 - accuracy: 0.4259\n",
      "Epoch 8/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6862 - accuracy: 0.4258\n",
      "Epoch 9/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6862 - accuracy: 0.4251\n",
      "Epoch 10/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6862 - accuracy: 0.4261\n",
      "Epoch 1/10\n",
      "1500/1500 [==============================] - 4s 2ms/step - loss: 0.6855 - accuracy: 0.4191\n",
      "Epoch 2/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6850 - accuracy: 0.4230\n",
      "Epoch 3/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6850 - accuracy: 0.4247\n",
      "Epoch 4/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6850 - accuracy: 0.4244\n",
      "Epoch 5/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6850 - accuracy: 0.4246\n",
      "Epoch 6/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6850 - accuracy: 0.4270\n",
      "Epoch 7/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6849 - accuracy: 0.4265\n",
      "Epoch 8/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6849 - accuracy: 0.4265\n",
      "Epoch 9/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6849 - accuracy: 0.4259\n",
      "Epoch 10/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6850 - accuracy: 0.4255\n",
      "Epoch 1/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6839 - accuracy: 0.4236\n",
      "Epoch 2/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6833 - accuracy: 0.4258\n",
      "Epoch 3/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6833 - accuracy: 0.4267\n",
      "Epoch 4/10\n",
      "1500/1500 [==============================] - 4s 2ms/step - loss: 0.6833 - accuracy: 0.4269\n",
      "Epoch 5/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6833 - accuracy: 0.4267\n",
      "Epoch 6/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6833 - accuracy: 0.4266\n",
      "Epoch 7/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6833 - accuracy: 0.4266\n",
      "Epoch 8/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6833 - accuracy: 0.4272\n",
      "Epoch 9/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6833 - accuracy: 0.4276\n",
      "Epoch 10/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6833 - accuracy: 0.4274\n",
      "Epoch 1/10\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1500/1500 [==============================] - 4s 2ms/step - loss: 0.6819 - accuracy: 0.4240\n",
      "Epoch 2/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6811 - accuracy: 0.4278\n",
      "Epoch 3/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6810 - accuracy: 0.4278\n",
      "Epoch 4/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6810 - accuracy: 0.4279\n",
      "Epoch 5/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6810 - accuracy: 0.4279\n",
      "Epoch 6/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6810 - accuracy: 0.4280\n",
      "Epoch 7/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6810 - accuracy: 0.4286\n",
      "Epoch 8/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6810 - accuracy: 0.4288\n",
      "Epoch 9/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6810 - accuracy: 0.4276\n",
      "Epoch 10/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6810 - accuracy: 0.4287\n",
      "Epoch 1/10\n",
      "1500/1500 [==============================] - 4s 2ms/step - loss: 0.6791 - accuracy: 0.4273\n",
      "Epoch 2/10\n",
      "1500/1500 [==============================] - 4s 2ms/step - loss: 0.6783 - accuracy: 0.4299\n",
      "Epoch 3/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6782 - accuracy: 0.4303\n",
      "Epoch 4/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6782 - accuracy: 0.4302\n",
      "Epoch 5/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6782 - accuracy: 0.4302\n",
      "Epoch 6/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6782 - accuracy: 0.4294\n",
      "Epoch 7/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6782 - accuracy: 0.4300\n",
      "Epoch 8/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6782 - accuracy: 0.4300\n",
      "Epoch 9/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6782 - accuracy: 0.4305\n",
      "Epoch 10/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6782 - accuracy: 0.4304\n",
      "Epoch 1/10\n",
      "1500/1500 [==============================] - 3s 2ms/step - loss: 0.6761 - accuracy: 0.4402\n",
      "Epoch 2/10\n",
      "1500/1500 [==============================] - 3s 2ms/step - loss: 0.6749 - accuracy: 0.4342\n",
      "Epoch 3/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6748 - accuracy: 0.4341\n",
      "Epoch 4/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6748 - accuracy: 0.4338\n",
      "Epoch 5/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6748 - accuracy: 0.4329\n",
      "Epoch 6/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6748 - accuracy: 0.4333\n",
      "Epoch 7/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6748 - accuracy: 0.4334\n",
      "Epoch 8/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6748 - accuracy: 0.4330\n",
      "Epoch 9/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6748 - accuracy: 0.4332\n",
      "Epoch 10/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6748 - accuracy: 0.4335\n",
      "Epoch 1/10\n",
      "1500/1500 [==============================] - 3s 2ms/step - loss: 0.6719 - accuracy: 0.4383\n",
      "Epoch 2/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6707 - accuracy: 0.4378\n",
      "Epoch 3/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6706 - accuracy: 0.4370\n",
      "Epoch 4/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6706 - accuracy: 0.4363\n",
      "Epoch 5/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6706 - accuracy: 0.4364\n",
      "Epoch 6/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6706 - accuracy: 0.4362\n",
      "Epoch 7/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6706 - accuracy: 0.4359\n",
      "Epoch 8/10\n",
      "1500/1500 [==============================] - 4s 2ms/step - loss: 0.6706 - accuracy: 0.4360\n",
      "Epoch 9/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6706 - accuracy: 0.4360\n",
      "Epoch 10/10\n",
      "1500/1500 [==============================] - 4s 3ms/step - loss: 0.6706 - accuracy: 0.4356\n"
     ]
    }
   ],
   "source": [
    "losses_width2 = []\n",
    "for lambda1 in np.linspace(-2,10,20):\n",
    "\n",
    "    xvals_1 = np.concatenate([w_true_alt,w_true])\n",
    "    yvals_1 = np.concatenate([np.ones(len(w_true_alt)),np.zeros(len(w_true))])\n",
    "    weights_1 = np.concatenate([np.ones(len(w_true_alt)),np.exp(lambda1*w_true**2)*len(w_true_alt)/np.sum(np.exp(lambda1*w_true**2))])\n",
    "\n",
    "    X_train_1, X_test_1, Y_train_1, Y_test_1, w_train_1, w_test_1 = train_test_split(xvals_1, yvals_1, weights_1)\n",
    "\n",
    "    Y_train_2 = np.stack((Y_train_1, w_train_1), axis=1)\n",
    "    Y_test_2 = np.stack((Y_test_1, w_test_1), axis=1)\n",
    "\n",
    "    inputs = Input((1, ))\n",
    "    hidden_layer_1 = Dense(50, activation='relu')(inputs)\n",
    "    hidden_layer_2 = Dense(50, activation='relu')(hidden_layer_1)\n",
    "    hidden_layer_3 = Dense(50, activation='relu')(hidden_layer_2)\n",
    "    outputs = Dense(1, activation='sigmoid')(hidden_layer_3)\n",
    "    model = Model(inputs=inputs, outputs=outputs)\n",
    "    \n",
    "    model.compile(loss=weighted_binary_crossentropy, optimizer='Adam', metrics=['accuracy'])\n",
    "    model.fit(X_train_1,\n",
    "              Y_train_2,\n",
    "              epochs=10,\n",
    "              batch_size=1000,\n",
    "              verbose=1)\n",
    "    losses_width2+=[model.history.history['loss'][-1]]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "id": "b10b748a",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.lines.Line2D at 0x7fc6184c0e20>"
      ]
     },
     "execution_count": 38,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(np.linspace(-2,10,20),losses_width)\n",
    "plt.axvline(2.5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "id": "ebde25dc",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x7fc628381d60>"
      ]
     },
     "execution_count": 39,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAaoAAAEBCAYAAAApVOHbAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjMuNCwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8QVMy6AAAACXBIWXMAAAsTAAALEwEAmpwYAAA1UklEQVR4nO3de5xV4/7A8c9Xuk6F0uVMIvUr5Raa6KKaoouj6UZKPxFlDjocyZ00HH6cjpEjlwwpDqmERnFU0hQKZwaHXJrjkiRUoiZdlL6/P9beY89uz8zee9aetfbM9/16zWs1az23vSz7O89az/MsUVWMMcYYvzrI6wYYY4wxZbFAZYwxxtcsUBljjPE1C1TGGGN8zQKVMcYYXzvY6wYkq8MPP1xbtWrldTOMMSapFBQUbFHVJrHksUAVp1atWpGfn+91M4wxJqmIyNex5rFbf8YYY3zNApUxxhhfs0BljDHG1yxQGWOM8TULVMYYY3zNApUxxhhfs+HpxlRD27dvZ9OmTezdu9frppgqombNmjRt2pSGDRu6XrYFKmOqme3bt/PDDz/QokUL6tati4h43SST5FSVXbt28e233wK4Hqzs1p8x1cymTZto0aIF9erVsyBlXCEi1KtXjxYtWrBp0ybXy7celd8tvzu6dL1vSmw7TJWxd+9e6tat63UzTBVUt27dhNxOth6VMdWQ9aRMIiTqurJAZYwxxtcsUBljjPE1e0ZljAFg6tJCT+qd0Leda2WtWbOGE044geXLl5Oenh5VnpycHJo2bcqQIUNca4dxl/WojDHVWk5ODgsWLPC6GaYMFqiMMcb4mgUqY0zSevjhh2nZsiUpKSlkZGTw3XfflTienZ1N586dOeSQQ2jWrBkZGRl8/vnnxcfT09MpKCjgySefREQQEWbNmgXAU089xemnn06jRo047LDD6N27t70s1SP2jMoYk5Ryc3MZP348l112GUOGDGHFihVccsklJdJs2LCBP//5zxx11FFs376d6dOn0717dwoLCznkkEN4+OGHOeecc2jdujWTJk0CoE2bNgCsW7eOCy+8kDZt2vDrr78ye/ZsevbsyZo1a2jdunWlf97qzAKVMSYp3XXXXQwYMIBHHnkEgP79+7N582Yef/zx4jRTp04t/vdvv/1G3759adq0Kbm5uVx44YUce+yxpKSk0KRJE7p06VKi/Ntuu6343/v376dv3778+9//5umnny5xzCSe3fozxiSd3377jffff5/BgweX2D9s2LASv7/99tv07duXxo0bc/DBB1OvXj127NhBYWH5Ixw//fRThg4dSrNmzahRowY1a9Zk7dq1UeU17rIelTEm6WzevJl9+/bRtGnTEvtDf1+/fj39+vXj1FNP5dFHHyU1NZVatWpx9tlns3v37jLLLyoqol+/fjRr1oz77ruPo446ijp16jBu3Lhy8xr3WaAyxiSdJk2acPDBBx+wAGro76+++io7d+4kNzeXlJQUAPbt28fWrVvLLX/16tVs2LCBpUuX0r59++L927Ztc+kTmFjYrT9jTNKpUaMGJ510Erm5uSX2v/DCC8X/3rVrFwcddBAHH/z73+Pz5s1j3759JfLUqlXrgF7Srl27AKhdu3bxvlWrVrFu3Tq3PoKJgfWojDFJ6eabb2bYsGFcfvnlDB06lBUrVvDqq68WH+/Tpw+//fYbF198MWPHjuXjjz/m3nvv5dBDDy1RTvv27Vm8eDGLFy+mcePGHH300XTp0oX69etz6aWXcv3117NhwwaysrJo0aJFJX9KAxaojDEBbi5lVBmGDh3KtGnTuOeee3jyySdJT09nxowZ9O/fH4ATTjiBmTNncvvtt/Piiy/SsWNHnnvuOUaMGFGinFtvvZX169dz3nnnsX37dmbOnMmYMWN47rnnuPbaaxk8eDBt27Zl+vTpTJkyxYuPWu2JqnrdhqSUlpamlTL5z95HZVz26aef0qFDB6+bYaqo8q4vESlQ1bRYyrRnVMYYY3zNApUxxhhfs0BljDHG1yxQGWOM8TULVMYYY3zNApUxxhhfs0BljDHG1yxQGWOM8TULVMYYY3zNApUxJillZWUhIrRt2zbi8f/5n/9BRMjKyiqx/4UXXqBPnz4ceuih1K5dm3bt2nHrrbeyZcuWSmi1iYet9WeMcUS7XJfbKrD8V506dfjqq6/Iz88nLe33VXn+/e9/8/XXX1OnTp0S6SdOnMj999/PxRdfzIQJE2jYsCGffPIJ06dP5+OPP+bFF1+Muy0mcSxQGWOSVkpKCqeccgpz5swpEajmzJlDnz59KCgoKN63cOFC7rvvPmbMmMEll1xSvL9Xr15kZmayZMmSSm27iZ7d+jPGJLWRI0cyb948ggtsqyrz5s1j5MiRJdJNnTqVU045pUSQCqpRowZnnXVWpbTXxM4ClTEmqQ0bNowffviBN998E4A33niDzZs3M3To0OI0e/fuZdWqVQwYMMCrZpoKsEBljElqhx56KAMGDGDOnDmAc9tvwIABJV6Q+OOPP7Jnzx6OPPJIj1ppKsIClTEm6Y0cOZL58+ezZ88e5s+ff8BtvyARqeSWGTdYoDLGJL1BgwaxY8cObrnlFn755RcyMjJKHG/cuDG1a9dm/fr1HrXQVIQFKmNM0ktJSWHgwIFMnTqVjIwMUlJSShyvWbMm3bt3Z/HixR610FREzIFKRHqIyPMi8p2I7Alsl4jIHyOk7SYir4jIVhHZKSIfisjVIlKjjPIvEpF3RWSHiGwTkTwRGVhG+roicruIrBWR3SKySUTmiUip70IWkSNE5AkR2Rj4DOtE5H4ROSzW82GM8YfLL7+cjIwMLrvssojHr776avLz83nyyScPOLZ//35effXVRDfRxCmmeVQicivwV2ALsAj4DjgcOBlIB14JSTsYeB7YDcwFtgIZwFSgOzA8Qvn3AhOBDcBjQC1gJLBQRK5U1QfD0tcGlgbKywf+AbQMlH22iPRR1XfC8rQBVgFNgVzgM+BU4C/AABHprqo/xnJejDHeS09PJz09vdTjGRkZXHPNNYwdO5a33nqLwYMHU79+fT777DOmT59Oq1atbFSgT0UdqERkOE6Qeg0YpqpFYcdrhvy7IU6g+Q1IV9X8wP5JwOvAuSIyUlXnhOTphhOkvgA6q+pPgf1/BwqAe0VkkaquC6n2GpwgNR8Yoar7A3nmAguAJ0TkhOD+gIdxgtRVqjotpP77gAnAXUDkP8mMqcoqsEJEssjOzqZbt248+OCDjBo1il27dtGqVSsGDRrEtdde63XzTCkkOEmuzEQiBwGfA82AVqq6uZz0lwAzgKdU9aKwY32AZcBKVe0Vsv8pYDRwiarODMtzBzAJuENVJwf2CbAOOBJorapfheVZCfQA+qjq8sC+1jiBcB3QJjSAiUgDnB6iAE1V9ZeyPmNaWprm5+eXlcQd0S5rUw2+ZIw7Pv30Uzp0KPXOuDEVUt71JSIFqppWaoIIon1G1Q04GufW3k8icraI3CAifxGRrhHS9wlsI930XQnsBLoFbt1Fk+dfYWkA2uAEqcLwIFVGnuC/l4T1sgj0EN8C6gFdIpTniaKde1i46lMybn4KgFF3zmH2sg8AkD43AzB72QeMGjXKqyYaY0xCRXvrr3Ng+wPwHnBC6MFA7+XckJ7WMYFtYXhBqrpPRL4CjgNaA5+KSArQAtihqt9FqP+/gW27kH2l1lHBPP0CeZaFHxSRTCATqLSJg6nD76bo5Swyujl/ocy+9ff5Ifr6/wEw6oyTGHWn9aiMMVVTtD2qpoHtZUBd4EygAXA8sBjoCTwXkv6QwHZbKeUF9x8aZ/rKzFNMVXNUNU1V05o0aVJKEe4qejkrqnTh80aMMaaqiDZQBYeTC07PaZmq7lDVj4GhOKP0epVyGzCS4PTw8h+QlRRL+njqiLddCZOz6N2o0mVmZia4JcYY441oA9VPge2Xqvqf0AOqugunVwXOMG/4vWdyCJE1DEtXXvpIPaFY64g3j6fy134bVbqyhuUaY0wyizZQrQ1sfy7leDCQ1Q1L3y48oYgcjDMwYx/wJUBghN23QH0R+UOE8oOv8Ax9tlRqHS7n8VTOxKHlJwJSU1MT3BJjjPFGtIFqJU5gaSsitSIcPz6wXRfYvh7YRpo91xNnZN0qVd0Tsr+sPGeFpQFnmPl6oJ2IHB1lnuWBbb/AkPtigeHp3YFdwNsRyvNE+oTHokpXVFRUfiJjjElCUQUqVd2Cs7rEIcBtocdEpC/QH+d2WXBo+Xyc1StGikhaSNo6wJ2BXx8Jq2Z6YHtL6FJGItIKGA/sAYrnV6kzASyYZ0po4AmsitED+ARYEZLnC2AJECwz1O1ACs7crzLnUFWmrIvOiCpdTk5OgltijDHeiGUJpWuA03ACSU/gXeAonMEUvwGXqurPAKq6XUQuxQlYeSIyB2cJpUE4Q8Tn4wS+Yqq6KrA6xDXAhyIyH2cJpRFAI+DKsFUpAO4DBgLnAu+IyDKcuVXDceZqXRI+Xwq4AmcJpQdE5Azg08Dn6o1zy++WGM5JwrU74vCo0uXn59uACmNMlRT1orSqugnnC30qznp6V+FMoH0Z6KGqz4WlXwD0wrlteA5wJbAXJxCN1AhLYqjqRGAM8D3OfKULgY+BjPB1/gLp9+AMlb8DZ0j5BKAvzvJJncPX+Qvk+QJIA2YFPs9EnMnDDwBd/bbOX9rlD0WVznpUxpiqKqbV01V1q6peo6pHq2otVW2sqoNVNeIzHVV9S1X/qKqHqWpdVT1BVaeq6m9l1PGkqnZW1RRVbaCqvVR1URnpd6nqZFVtq6q1VbWJqg5X1U/KyPONql6sqn8IfI6jVPUvqro1lvNRGTY+F91EXhv1Z9yQk5NT3DNPT08nLy+PjRs3Fg/Wyc7OZuLEiQB06tSJgoICCgsLadfOGZ+UlZVFVlYWAO3ataOwsJCCggI6deoEwMSJE8nOzgacAUAbN24kLy+vEj9h6datW4eIsGhRqV83EeXl5SEirFmzpsx08+bNY9asWRVoYUk7duxARMotU0QQEZ555pkDjv3zn/8sPh7uhx9+4Oqrr6ZNmzbUrl2bww47jLPOOsuTV6VEtdafOVBlrfWXffkfmXhej3LT5UlXC1YmKqWtxdagQQNPBuUEA6LX9uzZw/vvv0/79u1LvMa+PHl5efTu3ZuPPvqI448/vtR05557Llu2bHHts+7YsYMGDRowc+ZMxowZU2o6EaF+/fr06tXrgCB89tlns3LlSnbs2EFoLFi7di29e/cmJSWFa665hmOPPZbt27fzyiuv8MQTT/Duu+/SsWPHiPUlYq2/mF7zYSrfxh+j++II/kVrTLw2btzoSb1+CFIAtWvXpksX3yzz6aqMjAzmz5/PTz/9xGGHOWPVtm7dytKlSxk+fDizZ88ukf5///d/adSoEatWraJhw4Ylyrn88stjCuRusDf8+lz25Qe8jzKitLSY/kAx5gBeBYxYBwG9/vrriEiJwNq1a1dq1KjBzz//XLzvhBNO4JZbfh8btX79ekaOHEmjRo2oV68e/fv3Z+3atcXHI93627NnT/EXc+PGjbnuuuu4//77I94q27JlC8OHD6d+/fq0bt2ahx9+uPjYmDFjeP7551mxYkXxrbbgLVKA3Nxc0tLSqFOnDs2bN+f6669n7969Jcp//vnnadeuHXXr1qVnz5589tlnUZ+zrl27kpqayvPPP1+ivNTUVLp2Lbmg0MqVKykoKODuu+8uEaSCTjzxxEpb6zTIApXPdfrTAWNIIvLqr2FTdXg1ICfWP7K6du1KzZo1eeONNwDYuXMnBQUF1KpVi7feegtwegsff/wxPXr0KP799NNPZ+3atUyfPp158+bxyy+/cOaZZ7Jr165S67r++uuZNWsWkydP5plnnmH9+vXFz9jCXXrppXTs2JEXX3yR9PR0xo8fz7vvOkugTZo0id69e3PyySezevVqVq9ezbhx4wDn2dWwYcM49dRTeemll5g8eTI5OTncdNPvz6ffe+89RowYQceOHXnhhRcYNGgQ5513XtTnTEQYMWIEzz77bPG+Z599lpEjRx6QdsWKFdSoUYMzzzwz6vITTlXtJ46fTp06aWXInz5e9fX/K/fn3nvvrZT2mOT3ySefeN2ECuvSpYuOHz9eVVWXLVumjRs31hEjRugNN9ygqqq5ubl60EEH6bZt21RV9dZbb9VGjRrpjz/+WFzG1q1btWHDhvrggw+qqupXX32lgC5cuFBVVbds2aJ16tTRKVOmFOfZv3+/Hnvssep8dTqWL1+ugE6aNKl436+//qqHH354cXtUVc855xzt1atXic+xf/9+PfLII3XMmDEl9s+YMUPr1KmjW7ZsUVXV4cOHa4cOHXT//v3Fae68804FdObMmWWeK0CnTZum7733nh500EH63Xff6XfffacHHXSQvv/++zpt2rQSn+dPf/qTNm/evMwyy1Le9QXka4zft9aj8rkGdWuXnwjrUZmK8+qdZg0aNIg5T48ePYp7VCtXrqRHjx706tWrxL6OHTsW37p67bXX6Nu3Lw0bNmTfvn3s27ePBg0a0KlTJ0obFPXRRx+xe/duBg0aVLxPREp9U0G/fv2K/12zZk3atm3Lhg0byvwchYWFrF+/nvPOO6+4Xfv27aNPnz7s3r27eCThu+++y6BBg0rcchw2bFh5p6mEk08+mbZt2/Lcc88xb9482rZty0knnRQxbaRbm16yQOVzA295Mqp0pd2OMCZaAwcO9KTeeP7I6tmzJ2vWrOHnn3/mjTfeoEePHvTo0YP8/Hx2795dvC9oy5YtzJ07l5o1a5b4Wb58Od98803EOr7//nsAwl/pU9orfsIHGNSqVYvdu3eX+Tm2bNkCwB//+McS7Tr6aGdVuGDbvv/+e5o2bVoib/jv0RgxYgRz5sxhzpw5EW/7AbRo0YLNmzeX2/bKZKP+fK7wqYlRpQvOaTEmXl71qPLy8mJ+n1r37t1RVfLy8nj77bf529/+xnHHHUf9+vVZtmwZ7733Htddd11x+kaNGjFo0CAmTZp0QFml9eiaN28OwObNm2nUqFHx/s2bN0dMH49guTk5OZx88skHHA8GrObNm7Np06YSx8J/j8bIkSP561//CsATTzwRMU16ejq33XYby5Yt4+yzz465jkSwHpXPZc16Lap0tjKFqSivbvfEc+0edthhHH/88UydOpUaNWpw8sknIyKcfvrpTJkyhX379nH66acXpz/jjDP4+OOPOe6440hLSyvxc8wxx0Ss44QTTqBOnTrk5uYW71NVFi5cGPuHJHIP65hjjqFFixasW7fugHalpaXRuHFjADp37sxLL71UYq7TCy+8EHMbOnToQGZmJpmZmbRv3z5imh49etCpUyduvvnmiPPqPvroo1J7oYliPaoqosHHs2H7kugS97bX1psDhX4JVqZ4v/h79uzJQw89RP/+/alRw3m3a48ePbjuuuto27ZtcY8I4JprruHpp5+mT58+XHnllbRo0YIffviBFStWcPrpp3P++ecfUH7jxo259NJLmTx5MjVr1qRDhw7MnDmT7du3xxXU27dvT25uLgsWLOCII44gNTWV1NRUsrOzGT16NNu3b+ess86iVq1afPnllyxYsID58+dTr149brjhBk477TTOO+88xo4dy5o1a5gxY0Zc52369OnlpnnmmWfo3bs3aWlpTJgwoXjC7+LFi3nsscd45513aNmyZVz1x8N6VD6XNSa6IaLRPssypjTxLn8UXBElMzOzuHcUXOVi4cKFxbf1Ro0aVTyxNPhFP3v27LhvOQafQfXs2fOAfaG9KYDDDz+ct99+m/bt2zNhwgT69evH9ddfz7Zt2zjxxBNLrWPKlCmMGTOGrKwszj//fJo1a8bYsWMjzi8qzxVXXEG/fv245JJL6Ny5c/G5GjFiBLm5uXzwwQcMHz6cYcOG8fDDD3PKKadQq5bzVqW0tDTmzJnD+++/z5AhQ1iwYAFz584tq7oKOeaYY3jvvfcYMGAAU6ZMoW/fvowePZrCwkJmz55d6qoUiWJLKMWpspZQatfy8KifU0XNelTVWnlL3JiynXnmmezdu5cVK1aUn7gasiWUqqFFd10UVbqsWa9F3fsyxkRn+fLlvPPOO5xyyins3buXuXPnsmzZMp577rnyMxvXWKDyuaJde8pPZIxJiPr167NgwQLuvvtudu/eTdu2bZk1axbnnnuu102rVixQ+Vxm9osUPPrnctNZb8oY93Xu3Jm33474FiNTiSxQ+dyDN5zP6i/Lf5fjRVmz3H+WZYwxPmCj/nzugbl5UaWL9lmWMeDdUHRTtSXqurJA5XOHH1o/qnT2LMtEq2bNmmWuGG5MvHbt2kXNmjVdL9cClc+N6h/dKM7M7BcT3BJTVTRt2pRvv/2WnTt3Ws/KuEJV2blzJ99++21caxCWx55R+dygidN5KfuyctNFM+DCGKB4surGjRsPeDmfMfGqWbMmzZo1i2sydHksUPncjEkXRJVu4iOvRP02YGMaNmyYkC8UYxLBbv353Dff/xRVutTGsb/TxxhjkoEFKp+b8dKqqNJNPK9H+YmMMSYJWaDyuYeuHxFVutThdye4JcYY4w17RuVz9zy1hBsv7Fduuuk3jYpqYjBA194VbZUxxlQe61H5XPujmpefiOifZRljTLKxQOVzQ3qV/q6cUMFnWfc8tYQFKz4E4MzxD/DL7l9584MvuO4BZ57V5JyXi98JZIwxycDeRxWnynofVUqdWrz20FWultl17L2ulmeMMdGK531U1qPyudwoJvvGKp7XaBtjjFcsUPnc+59943qZ1os2xiQTC1Q+l7vyQ9fLtGdUxphkYoHK5/5+1VDXy1y0aJHrZRpjTKJYoPK5yTkvu16m9aiMMcnEApXPde/Y2vUyMzIyXC/TGGMSxQKVz/U7rYPrZWZmZrpepjHGJIoFKp/rNi7b9TLT09NdL9MYYxLFApXPrXp8outlpqamul6mMcYkigUqn1vyzqeul1lUVOR6mcYYkygWqHzurf986XqZOTk5rpdpjDGJYoHK527PPNv1MitjjUJjjHFL3IFKREaLiAZ+xpWSppuIvCIiW0Vkp4h8KCJXi0iNMsq9SETeFZEdIrJNRPJEZGAZ6euKyO0islZEdovIJhGZJyKlDpcTkSNE5AkR2Sgie0RknYjcLyKHxXYWEi+46rmbrEdljEkmcQUqEWkJTAN2lJFmMLAS6Am8CDwE1AKmAnNKyXMvMAv4A/AY8DRwArBQRP4cIX1tYClwG7Ad+AfwGjAUyBeR0yLkaQMUABcD7wba8yXwF2C1iDQu7/NXpsE9o3vNRyxs1J8xJpnEHKjEWXp7JvAjML2UNA1xAs1vQLqqjlXV64CTgNXAuSIyMixPN2Ai8AVwoqpOUNXxQCdgK3CviLQKq+oaoDswHzhNVW9Q1VHAuUA94AkRCf+MDwNNgatUdYiq3qiqfXAC1jHAXbGek0Q6uX1L18vMyspyvUxjjEmUeHpUVwF9cHokv5SS5lygCTBHVYsfiKjqbuDWwK+Xh+UJvs/iLlX9KSTPOpzeWO1AnUBxwAzmuV5V94fkyQXeAI4FeoXkaQ30A4Jlhpoc+DyjRSSllM9V6QZPjPi3QIW0a9fO9TKNMSZRYgpUgec+9wD/UNWVZSTtE9i+GuHYSmAn0C1w6y6aPP8KSwPQBjgSKFTVr6LME/z3ktDABqCqRcBbOD2xLhHK84TbL00ESEuL6Z1lxhjjqagDlYgcDPwTWA/cXE7yYwLbwvADqroP+Ao4GGgdKDsFaAHsUNXvIpT338A2tCtQah0u5/FU8LXybtq4caPrZRpjTKLE0qO6DTgZGKOqu8pJe0hgu62U48H9h8aZvjLzFBORTBHJF5H8zZs3l1KEuz77+nvXy8zOdn9ZJmOMSZSoApWInIrTi8pW1dUu1Bt8F3qsr5qNJX08dZSZR1VzVDVNVdOaNGkSQ7Hxu/HCfq6XaT0qY0wyKTdQhdzyKwQmRVlusGdySCnHG4alKy99pJ5QrHXEm8dT46fMdb1M61EZY5JJND2q+jjPbDoAu0Mm+SrOSDmAxwL77g/8vjawPeBZTyDwHQ3sw5m/hKr+AnwL1BeRP0RoQ9vANvTZUql1uJzHU2MHdXO9zE6dOrlepjHGJMrBUaTZA8wo5dgpOM+t3sQJAsHbgq8D/wsMAJ4Ny9MTZ2TdSlXdE7L/dWB0IM/MsDxnhaQJ+gJnYEc7ETk6wsi/SHmWB7b9ROSg0JF/ItIAZ07WLuDtiJ/WAy2bu79Yhq1MYYxJJuX2qFR1l6qOi/QDvBRI9mRgX/A+1XxgCzBSRIrHQotIHeDOwK+PhFUVnDB0S+hSRoFJvuNxAmZxAFNVDckzJXRib2BVjB7AJ8CKkDxfAEuAYJmhbgdSgKcCPTxfGPvXp10vs0GDBq6XaYwxiRJNjypmqrpdRC7FCVh5IjIHZ3WJQThDxOcDc8PyrBKR+3BWm/hQRObjLLk0AmgEXBmY/BvqPmAgzgTjd0RkGc7cquE4c7UuCZ8vBVwBrAIeEJEzgE+B04DeOLf8bqn4GXDPS9mXlZ8oRgMHDqSw0Dd3N40xpkwJWz1dVRfgrAqxEjgHuBLYixOIRgZ6ROF5JgJjgO+BTOBC4GMgQ1UfjJB+D3AmcAfOkPIJQF9gAdBZVd+JkOcLIA1nTcHTcJZtagM8AHRV1R/j/cyJMHux+yudW5AyxiQTiRAvTBTS0tK0Ml6XcX6/NK4ake5qmYu/qW/r/RljPCEiBaoa0/I49j4qn3M7SBljTLKxQOVzF9/xT9fLtN6UMSaZWKDyuRsSsDKFrZ5ujEkmFqh8rl6dmq6XuWjRItfLNMaYRLFA5XPXTVvgeplFRUWul2mMMYligcrn5t51ietlZmZmul6mMcYkSkIm/Br3PJ67inGD3V3v74J7nmXq0vLnUk3oa8+yjDHesx5VNZT76D1eN8EYY6Jmgcrn3O5NARzSuKnrZRpjTKJYoPK5Ebc84XqZ6ee6/9zLGGMSxQKVz/39yiGul5k18nTXyzTGmESxQOVzO3fvdb3MCQ8973qZxhiTKBaofO5vTy1xvczNG9a5XqYxxiSKDU/3uZm3jXa9zNWP3cwF14+IIuW9rtdtjDGxsh6Vzz0wN8/1Mh+KKkgZY4w/WKDyucMPre96mfck4HaiMcYkigUqnxvVP6b3i0Wl/VHNAThz/AP8svtX3vzgC6574EUAJue8zJJ3PgVARFyv2xhjYmVv+I1TZb3ht8mh9Xkp+7KE1xNJ17H2jMoY4y57w28VNGPSBZ7VPXv2bM/qNsaYIAtUPvfN9z95Vre9t8oY4wcWqHxuxkurPKvbelTGGD+wQOVzXg4lz8jI8KxuY4wJskDlc14OJbcXLBpj/MAClc8Fh5J7IT093bO6jTEmyAKVzw3pdaJndaempnpWtzHGBFmg8rkzxz/gWd1FRUWe1W2MMUEWqHwu16PJvgA5OTme1W2MMUEWqHzu/c++8azuylh5wxhjymOByudyV37oWd3WozLG+IEFKp/7+1VDPavbRv0ZY/zAXpzoc5NzXub2zLM9qfu4gWOZurQwqrQT+rZLcGuMMdWV9ah8rnvH1p7V3eSIVp7VbYwxQRaofK7faR08q3vq+HM8q9sYY4IsUPlct3HZntWdNedNz+o2xpggC1Q+t+rxiZ7VnTf/Cc/qNsaYIAtUPhd8LbwXtv24ybO6jTEmyAKVz731ny89q3vwn270rG5jjAmyQOVzXg1NB7jvimGe1W2MMUEWqHzuugde9Kzu4Vff4VndxhgTZIHK5wb39O41H7XrpXhWtzHGBFmg8rmT27f0rO4Zk7xbud0YY4KiClQi0lhExonIiyLyuYjsEpFtIvKmiIwVkYjliEg3EXlFRLaKyE4R+VBErhaRGmXUdZGIvCsiOwJ15InIwDLS1xWR20VkrYjsFpFNIjJPREqdKSsiR4jIEyKyUUT2iMg6EblfRA6L5nxUpsETp3tW900zF/PqU9N49alpANx9cX82bfiKbwrXFD+/yn30HhvGboxJKFHV8hOJXAY8AnwHLAfWA82AYcAhwPPAcA0pTEQGB/bvBuYCW4EM4BhgvqoOj1DPvcBEYAMwH6gFjAQaAVeq6oNh6WsDy4DuQD7wOtASGA78CvRR1XfC8rQBVgFNgVzgM+BUoDewFuiuqj+Wd07S0tK0Ml6DsXrGtQmvozRvH5kZddq/X5TOxo0bE9gaY0xVICIFqpoWU54oA1UfIAV4WVX3h+xvDryLExzOVdXnA/sbAp/jBLHuqpof2F8HJ5h0Bc5X1TkhZXUD3gK+ADqr6k+B/a2AgkD97VV1XUiem4D/wwlqI4JtCwTJBcAnwAlhbV4M9AOuUtVpIfvvAyYAj6pqufe8KitQ3XBhP09fRx+to866xl5db4wpVzyBKqpbf6r6uqouDP3CD+z/Hgjem0oPOXQu0ASYEwxSgfS7gVsDv14eVk0wONwVDFKBPOuAh4DawMXB/SIiIXmuD22bquYCbwDHAr1C8rTGCVLBMkNNBn4BRouIb0YRfPb19143ISqFhdGtsm6MMbFyYzDF3sB2X8i+PoHtqxHSrwR2At0Ct+6iyfOvsDQAbYAjgUJV/SrKPMF/L4kQdItwenT1gC4RyvPEjRf287oJUcnKyvK6CcaYKqpCgUpEDgYuDPwaGmCOCWwP+DNbVfcBX+G8C6t1oJwUoAWwQ1W/i1DVfwPb0JcelVqHy3mKiUimiOSLSP7mzZtLKcJd46fMrZR6KiovL8/rJhhjqqiK9qjuAY4HXlHVxSH7Dwlst5WSL7j/0DjTV2aeYqqao6ppqprWpEmTUopw19hB3SqlnorKzIx+4IUxxsQi7jf8ishVOCP0PgNGx5o9sC1/JEdJsaSPp45425UwLZv7bsR8RD/XaxnV24DtTcDGmFjF1aMSkfHAP3BG1fVW1a1hSYI9k0OIrGFYuvLSR+oJxVpHvHk8NfavT3vdhKh0PXuE100wxlRRMQcqEbkaeBBYgxOkIg1LWxvYHvDnc+C51tE4gy++BFDVX4Bvgfoi8ocI5bUNbEP/ZC+1DpfzeOql7ORYHeKmwSd73QRjTBUVU6ASkRuAqcAHOEGqtBcWvR7YDohwrCfOyLpVqronyjxnhaUBZ77VeqCdiBwdZZ7lgW2/8NU0RKQBzsThXcDbEcrzxOzFiZ+r5YbJz77hdROMMVVU1IFKRCbhDJ4oAM5Q1S1lJJ8PbAFGikjxxK7AhN87A78+EpYnOB/rltCljAITfscDe4CZwf2BVTCCeaaEBp7AhN8eOLcmV4Tk+QJYAgTLDHU7zqTipwI9PF/Y8vMOr5sQlS/+867XTTDGVFFRDaYQkYuAO4DfcCbSXuXMty1hnarOAlDV7SJyKU7AyhOROThLKA0isIQSzrJKxVR1VWB1iGuAD0UkuITSCH5fQmldWJ33AQNxJhi/IyLLcOZWDceZq3VJ+Hwp4AqcJZQeEJEzgE+B03CWUCoEbonmnFSWq0ake92EqKx+ZR7Hde1TfkJjjIlRtEsoZeGs3FCWFaqaHpavO84Xf1egDs6ySk8AD6jqb6XUdRHwZ5xVJfYD7wF/V9VFpaSvC9wIjMIJUtuBPGCyqn5SSp6WOIF3ANAYZw3DBcDtEQaGRFRZSyi1P6oZM2+LdVBl5Yt2XUAb9WdM9Zawtf7MgSorUM2cNJr2rZolvJ6KGv/MWkbflF1uOgtUxlRvCVvrz3inXp2aXjchKseelg7ANf2cxT8KXl/IP++eCMDjky7j49Wvs3vnDho0aOBVE40xScp6VHGqrB7Vkc0bMfeuSxJeT0XZrT9jTDSsR1UFJUOQikVOTo7XTTDGJBkLVD73eO4qr5vgqsrohRpjqpa41/ozFbD8bq9b4JkOw6+Nak1AsNuExhiH9ah8btzg5Fg9PVoPXev/ofbGGH+xHpXPjbjliaR4TtVlfXTPnj4f/ecEt8QYU9VYj8rn/n7lEK+b4KomR7TyugnGmCRjgcrndu7e63UTXDV1/DleN8EYk2QsUPnc355a4nUTXJU1503y5j9B7qP3AHDfFcP4pnANmzZ8xd0X9wfg1aem8epT07xspjHGR2zCb5wqNOE3hlF/q7/8Mb46fCraicEAj4wfSGGhb14NZoxxgU34rYIemJvndRM8s2hRxHWIjTHVjAUqnzv80PpeN8EzRUVFXjfBGOMDFqh8blT/mHrIVUpmZvS3CY0xVZc9o4pTRZ5RZXTtQObAzqSf1JrU4XdT9HIWOYveJX/tt+RMHEr6hMfIuugM2h1xOB0vncZL2Ze53HrvxPKMKlq2goUxycPeR1WJbDBFfGIJVLmP3sPgP91YbjoLVMYkDxtMkSRG3TnH6yYkhUMaNwUga+TpbPvxBz7/zzvFSzDNmzqJ1S/PBbB3XBlTxVmPKk4V6VHdnnk2/U7r4HKLqp5oe1/juvzBgpUxScJ6VEnCgpS78vLyvG6CMSaBLFB5oNu4bK+bUKXYyxiNqdosUHlg1eMTvW5ClbJw4UKvm2CMSSALVB5Y8s6nXjehSjmlTwYX3JTN1KWFiAhTlxZywU3ZnNIng6lLCzmuSx/G3TGdu3Pfs2dZxiQhG0wRp4oMpuh3Wgduzzzb5RZVPTbnypiqxwZTJAkLUt6x51nGJB97w68HrnvgRf5+1VCvm+F70b41OJaeV35+vi3NZEySsR6VBwb3PNHrJlRbOTk5pKenk5eXx8aNG0lNTQUgOzubiROdQS6dOnWioKDAy2YaY0LYM6o4VeQZ1WsPXUVKnVout6j6SsSzrLOPgnbt7HmWMW6zZ1RJYvDE6V43wZRj4MCBXjfBGBNggcoDrz10lddNMOWwNwsb4x82mMIDC1Z8yJBe9pzKLdEOuoDobxNmZWUVb9u1a8eiRYsoKioiMzOTgoICJk6cSGpqavF248aN8TTdGBMFC1Qe+Ozr7wELVH4WDFRQsncVHGSRnf37Mlhxv+7FGBMVC1QeuPHCfl43wZRj6tLob/2dXPP30YPGGPfZMyoPjJ8y1+smGBcFe1+ZmZnFE4obNGhAUVERCxcuJCMjA4BRo0Z51URjkpoNT49TRYanP3TdCE5p39LlFhk3xTLkPdplmWbPnm3BylR78QxPt1t/HmjZ/DCvm2BcFO1twgmjRiEiqCqzZ89m0aJFzJ49m4yMDDIzM0lPTyc1NZWioiJycnJsBQ1jAuzWnwfG/vVpr5tgPBK8gzFq1Chmz54NUHx7MHi7EGyAhjGhrEflgZeyL/O6CaYciRjyHgtbPNeY31mg8sDsxfmM6h/TLVpTBcQykjD3rsziOVxpaWls3LiR7Ozs4m2nTp3IycmhQYMGttSTqfIsUHlgy887vG6CcVEiVnnPy8sr/ndwMnFw0Vz4fT5Xu3btbBUNU+XZMyoPXDUi3esmmCrCgpSpDqrt8HQROQK4AxgANAa+AxYAt6vqT+Xlr8jw9PZHNWPmbaPjymuqh2h7X9vecgZk2FJPJlnEMzy9WgYqEWkDrAKaArnAZ8CpQG9gLdBdVX8sq4yKBKqZk0bTvlWzuPKa6iHaQBXtHC6gxPu3jPGKzaOK3sM4QeoqVZ0W3Cki9wETgLuAhA3Nq1enZqKKNlVEtM+9pi6N/rmXLfVkklW161GJSGvgC2Ad0EZV94cca4BzC1CApqr6S2nlVKRHdWTzRsy965K48hoTr5v+mV9ikIYxXrAXJ0anT2C7JDRIAahqEfAWUA/okqgGWJAyXhh8Sw5d/ziC867+K1OXFlKnXgp3577HuDumc1yXPkxdWsgpfTKKJyIb4xfVsUf1d+Ba4FpVzY5w/EFgPHCFqj5SWjkV6VGNHdSNcYO7xZXXmER7+8hMrul3DPctWUvB6wv55J08Rt+UzeOTLqPrH8+jTcdTuf38Htyd+z4pX+XZUk8mJjaYIgoikgNcClyqqo9HOH4XcDNws6reHXYsEwj+X3kMzsCLeBwObIkzb3Vk5ys2dr5iY+crdhU5Z0epapNYMlTXwRRlkcD2gAiuqjlAhde2EZH8WP+iqM7sfMXGzlds7HzFrrLPWXV8RrUtsD2klOMNw9IZY4zxUHUMVMHbdaVNQGkb2NqUf2OM8YHqGKiWB7b9RKTE5w8MT+8O7ALeTmAbbGns2Nj5io2dr9jY+YpdpZ6zajeYAkBEFgP9KH3C76Oqau/iMMYYH6iugSp8CaVPgdNwllAqBLqVt4SSMcaYylEtAxWAiLSk9EVpt3rYNGOMMSGqbaAyxhiTHKrjYIq4iMgRIvKEiGwUkT0isk5E7heRwxJdjoh0E5FXRGSriOwUkQ9F5GoRqVHxT5YYFT1fItJYRMaJyIsi8rmI7BKRbSLypoiMDR8IE8jTSkS0jJ857n9Sd7hxfQXylPbZvy8jX9JdX+DKNTamnOtFReS3sDxJd42JyLkiMk1E3hCR7YF2Ph1nWZ58f1mPKgoRnmnF/FqQeMsRkcHA88BuYC6wFcjAWRljvqoOd+EjusqN8yUilwGP4NySXQ6sB5oBw3DmwD0PDNeQC1hEWgFfAf/BuY0bbo2qzq/AR0sIF6+vdcChwP0RDu9Q1Xsj5Em66wtcu8ZOAoaUcrgHzrqgL6vqwJA8rUiya0xEPgA6AjuADUB74BlVvSDGcrz7/lJV+ynnB1iMs1LFlWH77wvsn56IcnAmH28C9gBpIfvrBC4YBUZ6fX4Scb5wviQygIPC9jfHCVoKnBN2rFVg/yyvz4FH19c6YF0M9Sbl9eXmOSuj/NWBcgYl+zWGE0ja4qy6kx5o/9OJPuduXl+en0S//wCtAyf0qwhfmg1w/kr5BUhxuxzgkkCeJyOU1ydwbIXX5ygR56ucOm4O1DEtbH8yfom4dr7iCFRJd31VxjUGHB8ofwNQI9mvsbD2xxWovP7+smdU5XPrtSDxlBPM82qE8lYCO4FuIlK7vA9RiSrjNSp7A9t9pRxPFZE/icjNge2JFagr0dw+X7VF5ILAZ/+LiPQu41lAMl5fkPhr7E+B7QxV/a2UNMl0jbnB0+8vC1TlOyawLW1Jpf8GtuW9EzyeckrNo6r7cP66ORjnrx2/cOt8RSQiBwMXBn6N9D8AQF9gOs6bmqcD/xGR5SJyZDx1Jpjb56s58E+cz34/8DrwXxHpFUvdPr6+IIHXmIjUBS4A9gMHvF0hRDJdY27w9PvLAlX5govXlrZIbXD/oQkox626K1Oi23wPzq2ZV1R1cdixncBfgU7AYYGfXjiDMdKBZSKSEme9ieLm+ZoJnIETrFKAE4BHcW5X/UtEOiaw7sqUyHafF8j3L1X9JsLxZLzG3ODp95cFqoor9bUglVCOW3VXprjbLCJXARNxRhuNDj+uqptU9TZVfU9Vfw78rMRZLusd4H+AcfE33RNRny9VvV1VX1fVH1R1p6quUWcpsPuAukBWour2mYq0O/i+uUcjHayi15gbEvr9ZYGqfG69FiSecpLxlSQJabOIjAf+AXwC9NYYVg8J3GYI3sbpGUu9laAy/htPD2zDP3syXl+QuGvsWKAbziCKV2LJ6/NrzA2efn9ZoCqfW68FiaecUvMEntUcjTOg4Mty6q5Mrr9GRUSuBh4E1uAEqVInr5Zhc2Drt9sylfHamU2BbfhnT8brCxJ3zqIZRFEWv15jbvD0+8sCVfmWB7YVfS1IPOW8HtgOiFBeT5xRNqtUdU95H6ISuXW+gnluAKYCH+AEqU1l5yhVcDSS3750XT1fpega2IZ/9mS8viAB50xE6uDcTt4PzIizXX69xtzg7feX1+P6k+GHGCa6ATVxZn63qUg5gf0Ncf5KS6oJmS6er0mB9PlAoyjqPQ2oFWF/H5yZ8YqzMr7n58jt8wUcF+kcAUfhjMhS4OaqcH25eY2FpBkdyLewKl5jIe1Mp4x5VH79/rIllKIQYemQUl8LErLEyteq2ireckLyDAHm4/xPMAdnCZJBBJYgAc5Tn/1HdON8ichFwCzgN2Aake9jr1PVWSF58nC+sPNwnjMAnMjv8zkmqeqdFf6ALnPpfGUBN+L85fsVUAS0Ac7G+WJ4BRiqqr+G1T2EJLu+wL3/J0PKewM4HWclioVl1JtHkl1jgf/GQwK/Ngf64/T63gjs26Kq1wbStsKP319eR/hk+QFa4gz//Q74Ffga5+F+o7B0rXD+UlhXkXLC8nTH+aL5Cad7/RHOCx5ruPX5/Ha+cEaoaTk/eWF5xgKLcFZo2IHzl9x6nDXGenh9ThJ8vnoBz+KMiPwZZ1L0ZmApzrwzqUrXlxvnLOR4h8Dxb8r7zMl4jUXx/9K6kLS+/P6yHpUxxhhfs8EUxhhjfM0ClTHGGF+zQGWMMcbXLFAZY4zxNQtUxhhjfM0ClTHGGF+zQGWMMcbXLFAZY4zxNQtUxhhjfO3/ATtnJocorPWSAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "lambda1 = 2.5\n",
    "\n",
    "xvals_1 = np.concatenate([w_true_alt,w_true])\n",
    "yvals_1 = np.concatenate([np.ones(len(w_true_alt)),np.zeros(len(w_true))])\n",
    "weights_1 = np.concatenate([np.ones(len(w_true_alt)),np.exp(lambda1*w_true)*len(w_true_alt)/np.sum(np.exp(lambda1*w_true))])\n",
    "\n",
    "X_train_1, X_test_1, Y_train_1, Y_test_1, w_train_1, w_test_1 = train_test_split(xvals_1, yvals_1, weights_1)\n",
    "\n",
    "_,_,_=plt.hist(X_test_1[Y_test_1==1],bins=np.linspace(0,1,30),alpha=0.5,label=\"data\")\n",
    "_,_,_=plt.hist(X_test_1[Y_test_1==0],bins=np.linspace(0,1,30),alpha=0.5,label=\"MC\")\n",
    "_,_,_=plt.hist(X_test_1[Y_test_1==0],bins=np.linspace(0,1,30),weights=w_test_1[Y_test_1==0],histtype=\"step\",color=\"black\",ls=\":\",label=\"weighted MC\")\n",
    "plt.legend(fontsize=15)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "id": "34735921",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "-1.0 0.1496976693057114 0.12275355685389264 0.11445316799591017\n",
      "-0.42105263157894735 0.14988903868969533 0.1226426242190147 0.11897827524389355\n",
      "0.1578947368421053 0.14983003812970666 0.12271762324848029 0.1241585131932004\n",
      "0.736842105263158 0.15015646094778218 0.12283379456967299 0.12986512968452166\n",
      "1.3157894736842106 0.14962530021582787 0.1230320887574309 0.13616017754253884\n",
      "1.8947368421052633 0.15000024928847014 0.12301975922904755 0.14288509282901374\n",
      "2.473684210526316 0.1499887184124594 0.12264129382952356 0.14972190518140568\n",
      "3.052631578947368 0.14994553153594128 0.1229474788354419 0.15832508546373011\n",
      "3.6315789473684212 0.14976147401060763 0.122971830856749 0.16669621244819155\n",
      "4.210526315789474 0.14964570134502397 0.12268927338827264 0.17545532230078378\n",
      "4.7894736842105265 0.15011575952615983 0.12285439062032003 0.1857899403621811\n",
      "5.368421052631579 0.14959159961295193 0.12328623245028235 0.19771046656917665\n",
      "5.947368421052632 0.14963372114734327 0.12307227360738325 0.20868116895546773\n",
      "6.526315789473685 0.150213439490434 0.12283973173167455 0.21959380452118638\n",
      "7.105263157894736 0.14986528838016447 0.12274135251920833 0.23139333994163086\n",
      "7.684210526315789 0.14975467412372187 0.12255353019435236 0.24494789055520522\n",
      "8.263157894736842 0.14976876388335272 0.12284789999662851 0.2590379676452482\n",
      "8.842105263157896 0.15007360798457237 0.12282999344405185 0.2712923787451337\n",
      "9.421052631578949 0.1497036339289114 0.1230811234957034 0.2845341912482289\n",
      "10.0 0.14967715587019773 0.1229835851001833 0.29877883761503266\n"
     ]
    }
   ],
   "source": [
    "for lambda1 in np.linspace(-1,10,20):\n",
    "\n",
    "    xvals_1 = np.concatenate([w_true_alt,w_true])\n",
    "    yvals_1 = np.concatenate([np.ones(len(w_true_alt)),np.zeros(len(w_true))])\n",
    "    weights_1 = np.concatenate([np.ones(len(w_true_alt)),np.exp(lambda1*w_true)*len(w_true_alt)/np.sum(np.exp(lambda1*w_true))])\n",
    "\n",
    "    X_train_1, X_test_1, Y_train_1, Y_test_1, w_train_1, w_test_1 = train_test_split(xvals_1, yvals_1, weights_1)\n",
    "\n",
    "    print(lambda1,np.mean(X_test_1[Y_test_1==1]),np.mean(X_test_1[Y_test_1==0]),np.average(X_test_1[Y_test_1==0],weights=w_test_1[Y_test_1==0]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "id": "b837cd66",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "-1.0 0.03360062406006171 0.024149244284022982 0.022625021750417924\n",
      "-0.42105263157894735 0.033838311022465754 0.024090373933198555 0.02343281776930534\n",
      "0.1578947368421053 0.03359903915747429 0.0240870523306639 0.0243455240172513\n",
      "0.736842105263158 0.033782630367368505 0.024077798469549414 0.025319404881640094\n",
      "1.3157894736842106 0.03371316728182848 0.02410469132183544 0.0264073510788454\n",
      "1.8947368421052633 0.03374289894411669 0.024212950715400923 0.0277030038840649\n",
      "2.473684210526316 0.03350453230219468 0.024096205932296923 0.028801122841679294\n",
      "3.052631578947368 0.03382393314674943 0.02392242235593586 0.029841826632522467\n",
      "3.6315789473684212 0.033543882365101294 0.02411414677082154 0.03161374050070759\n",
      "4.210526315789474 0.03363742572332272 0.024144655383687288 0.03322569378860751\n",
      "4.7894736842105265 0.03377921273255536 0.024054969930184435 0.03473957696647103\n",
      "5.368421052631579 0.03359624568769006 0.02420171205697464 0.036856224575223556\n",
      "5.947368421052632 0.03377876132624582 0.024212233267877456 0.03876619600089746\n",
      "6.526315789473685 0.03374770741189845 0.0240376533519754 0.04057796489186755\n",
      "7.105263157894736 0.03372213177699258 0.024113565550687114 0.04317369188395326\n",
      "7.684210526315789 0.03354447759273202 0.024084129341254777 0.04545430452573111\n",
      "8.263157894736842 0.033582907226390665 0.024071346407654508 0.0480979151075097\n",
      "8.842105263157896 0.03381341970062108 0.02397824969154789 0.05047400823103534\n",
      "9.421052631578949 0.03382460392774575 0.024026048124735293 0.053761181567883044\n",
      "10.0 0.03390778003115195 0.024149241242945287 0.058171788622852015\n"
     ]
    }
   ],
   "source": [
    "for lambda1 in np.linspace(-1,10,20):\n",
    "\n",
    "    xvals_1 = np.concatenate([w_true_alt,w_true])\n",
    "    yvals_1 = np.concatenate([np.ones(len(w_true_alt)),np.zeros(len(w_true))])\n",
    "    weights_1 = np.concatenate([np.ones(len(w_true_alt)),np.exp(lambda1*w_true**2)*len(w_true_alt)/np.sum(np.exp(lambda1*w_true**2))])\n",
    "\n",
    "    X_train_1, X_test_1, Y_train_1, Y_test_1, w_train_1, w_test_1 = train_test_split(xvals_1, yvals_1, weights_1)\n",
    "\n",
    "    print(lambda1,np.mean(X_test_1[Y_test_1==1]**2),np.mean(X_test_1[Y_test_1==0]**2),np.average(X_test_1[Y_test_1==0]**2,weights=w_test_1[Y_test_1==0]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "id": "48e4a194",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0.65, 0.7)"
      ]
     },
     "execution_count": 42,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "losses_width = np.array(losses_width)\n",
    "losses_width2 = np.array(losses_width2)\n",
    "plt.plot(np.linspace(-2,10,20),losses_width)\n",
    "plt.plot(np.linspace(-2,10,20),losses_width2)\n",
    "plt.axvline(2.5,color=\"tab:blue\",ls=\":\")\n",
    "plt.axvline(4.5,color=\"tab:orange\",ls=\":\")\n",
    "plt.ylim([0.65,0.7])"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "efea46d4",
   "metadata": {},
   "source": [
    "## Try to do it in one step instead of scanning?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 90,
   "id": "67914c60",
   "metadata": {},
   "outputs": [],
   "source": [
    "from tensorflow.keras.layers import Layer\n",
    "from tensorflow.keras.layers import concatenate\n",
    "\n",
    "initializer = tf.keras.initializers.RandomUniform(minval=-5., maxval=5.)\n",
    "\n",
    "class MyLayer(Layer):\n",
    "\n",
    "    def __init__(self, myc, **kwargs):\n",
    "        self.myinit = myc\n",
    "        super(MyLayer, self).__init__(**kwargs)\n",
    "\n",
    "    def build(self, input_shape):\n",
    "        # Create a trainable weight variable for this layer.\n",
    "        self._lambda0 = self.add_weight(name='lambda0', \n",
    "                                    shape=(1,),\n",
    "                                    initializer=tf.keras.initializers.Constant(self.myinit), \n",
    "                                    trainable=True)\n",
    "        self._lambda1 = self.add_weight(name='lambda1', \n",
    "                                    shape=(1,),\n",
    "                                    initializer=tf.keras.initializers.Constant(self.myinit), \n",
    "                                    trainable=True)\n",
    "        super(MyLayer, self).build(input_shape)  # Be sure to call this at the end\n",
    "\n",
    "    def call(self, x):\n",
    "        #return tf.exp(self._lambda1 * x + self._lambda0)\n",
    "        return tf.exp(self._lambda0 * x + self._lambda1 * x**2)\n",
    "\n",
    "myc = 0.1\n",
    "mymodel_inputtest = Input(shape=(1,))\n",
    "mymodel_test = MyLayer(myc)(mymodel_inputtest)\n",
    "model_generator = Model(mymodel_inputtest, mymodel_test)\n",
    "\n",
    "inputs_disc = Input((1, ))\n",
    "hidden_layer_1_disc = Dense(50, activation='relu')(inputs_disc)\n",
    "hidden_layer_2_disc = Dense(50, activation='relu')(hidden_layer_1_disc)\n",
    "hidden_layer_3_disc = Dense(50, activation='relu')(hidden_layer_2_disc)\n",
    "outputs_disc = Dense(1, activation='sigmoid')(hidden_layer_3_disc)\n",
    "model_discrimantor = Model(inputs=inputs_disc, outputs=outputs_disc)\n",
    "\n",
    "def weighted_binary_crossentropy(y_true, y_pred):\n",
    "    weights = tf.gather(y_true, [1], axis=1) # event weights\n",
    "    y_true = tf.gather(y_true, [0], axis=1) # actual y_true for loss\n",
    "    \n",
    "    weights_1 = K.sum(y_true*weights)\n",
    "    weights_0 = K.sum((1-y_true)*weights)\n",
    "    \n",
    "    # Clip the prediction value to prevent NaN's and Inf's\n",
    "    epsilon = K.epsilon()\n",
    "    y_pred = K.clip(y_pred, epsilon, 1. - epsilon)\n",
    "    t_loss = -weights * ((y_true) * K.log(y_pred)/weights_1 +\n",
    "                         (1 - y_true) * K.log(1 - y_pred)/weights_0)\n",
    "    return K.mean(t_loss)\n",
    "\n",
    "model_discrimantor.compile(loss=weighted_binary_crossentropy, optimizer='adam')\n",
    "\n",
    "def weighted_binary_crossentropy_GAN(y_true, y_pred):\n",
    "    weights = tf.gather(y_pred, [1], axis=1) # event weights\n",
    "    y_pred = tf.gather(y_pred, [0], axis=1) # actual y_pred for loss\n",
    "    \n",
    "    weights_1 = K.sum(y_true*weights)\n",
    "    weights_0 = K.sum((1-y_true)*weights)\n",
    "    \n",
    "    #tf.print(\"weights\",weights_0,weights_1)\n",
    "    \n",
    "    # Clip the prediction value to prevent NaN's and Inf's\n",
    "    epsilon = K.epsilon()\n",
    "    y_pred = K.clip(y_pred, epsilon, 1. - epsilon)\n",
    "    t_loss = weights * ((1 - y_true) * K.log(1 - y_pred)/weights_0)\n",
    "    return K.mean(t_loss)\n",
    "    \n",
    "model_discrimantor.trainable = False\n",
    "mymodel_gan = Input(shape=(1,))\n",
    "gan_model = Model(inputs=mymodel_gan,outputs=concatenate([model_discrimantor(mymodel_gan),model_generator(mymodel_gan)]))\n",
    "\n",
    "gan_model.compile(loss=weighted_binary_crossentropy_GAN, optimizer='adam')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 95,
   "id": "313309fe",
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "on epoch= 0 [[6.25792]] [[-8.693584]]\n",
      "on epoch= 1 [[6.269087]] [[-8.856627]]\n",
      "on epoch= 2 [[6.309806]] [[-8.952942]]\n",
      "on epoch= 3 [[6.3436236]] [[-9.033758]]\n",
      "on epoch= 4 [[6.372823]] [[-9.090017]]\n",
      "on epoch= 5 [[6.3842044]] [[-9.143791]]\n",
      "on epoch= 6 [[6.4028273]] [[-9.175528]]\n",
      "on epoch= 7 [[6.4204044]] [[-9.201631]]\n",
      "on epoch= 8 [[6.432082]] [[-9.211593]]\n",
      "on epoch= 9 [[6.432207]] [[-9.229482]]\n",
      "on epoch= 10 [[6.433632]] [[-9.243059]]\n",
      "on epoch= 11 [[6.447183]] [[-9.251026]]\n",
      "on epoch= 12 [[6.4487634]] [[-9.262688]]\n",
      "on epoch= 13 [[6.454585]] [[-9.266682]]\n",
      "on epoch= 14 [[6.458421]] [[-9.270363]]\n",
      "on epoch= 15 [[6.458276]] [[-9.274342]]\n",
      "on epoch= 16 [[6.4543085]] [[-9.2809105]]\n",
      "on epoch= 17 [[6.4604273]] [[-9.276051]]\n",
      "on epoch= 18 [[6.4579544]] [[-9.273301]]\n",
      "on epoch= 19 [[6.4621825]] [[-9.267141]]\n"
     ]
    }
   ],
   "source": [
    "#xvals_1 = np.concatenate([gauss_data,gauss_sim])\n",
    "#yvals_1 = np.concatenate([np.ones(len(gauss_data)),np.zeros(len(gauss_sim))])\n",
    "\n",
    "xvals_1 = np.concatenate([w_true_alt,w_true])\n",
    "yvals_1 = np.concatenate([np.ones(len(w_true_alt)),np.zeros(len(w_true))])\n",
    "\n",
    "X_train_1, X_test_1, Y_train_1, Y_test_1 = train_test_split(xvals_1, yvals_1)\n",
    "\n",
    "n_epochs = 20\n",
    "n_batch = 128*10\n",
    "n_batches = len(X_train_1) // n_batch\n",
    "\n",
    "for i in range(n_epochs):\n",
    "    lambdasum = np.log(model_generator.predict([1.]))\n",
    "    lambdasum2 = np.log(model_generator.predict([2.]))\n",
    "    mylambda1 = (lambdasum2-2*lambdasum)/2\n",
    "    mylambda0 = lambdasum - mylambda1\n",
    "    print(\"on epoch=\",i,mylambda0,mylambda1)\n",
    "    #print(\"  \",np.sum(model_generator.predict(X_train_1,batch_size=1000)))\n",
    "    for j in range(n_batches):\n",
    "        X_batch = X_train_1[j*n_batch:(j+1)*n_batch]\n",
    "        Y_batch = Y_train_1[j*n_batch:(j+1)*n_batch]\n",
    "        W_batch = model_generator(X_batch)\n",
    "        W_batch = np.array(W_batch).flatten()\n",
    "        W_batch[Y_batch==1] = 1\n",
    "        #W_batch[Y_batch==0] = 1\n",
    "        \n",
    "        Y_batch_2 = np.stack((Y_batch, W_batch), axis=1)\n",
    "        \n",
    "        model_discrimantor.train_on_batch(X_batch, Y_batch_2)\n",
    "        \n",
    "        #print(\"      \",j,np.sum(model_generator.predict(X_batch,batch_size=1000)),np.log(model_generator.predict([1.]))-np.log(model_generator.predict([0.])),np.log(model_generator.predict([0.])))\n",
    "        \n",
    "        gan_model.train_on_batch(X_batch[Y_batch==0],np.zeros(len(X_batch[Y_batch==0])))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 96,
   "id": "c3be531a",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.03367921811165968 0.024050254362818498 0.03361572827524153\n",
      "0.14972844300053106 0.12275832281394203 0.14970402631204266\n"
     ]
    }
   ],
   "source": [
    "lambda0 = 6.5\n",
    "lambda1 = -9.3\n",
    "xvals_1 = np.concatenate([w_true_alt,w_true])\n",
    "yvals_1 = np.concatenate([np.ones(len(w_true_alt)),np.zeros(len(w_true))])\n",
    "weights_1 = np.concatenate([np.ones(len(w_true_alt)),np.exp(lambda1*w_true**2+lambda0*w_true)*len(w_true_alt)/np.sum(np.exp(lambda1*w_true**2+lambda0*w_true))])\n",
    "\n",
    "X_train_1, X_test_1, Y_train_1, Y_test_1, w_train_1, w_test_1 = train_test_split(xvals_1, yvals_1, weights_1)\n",
    "\n",
    "print(np.mean(X_test_1[Y_test_1==1]**2),np.mean(X_test_1[Y_test_1==0]**2),np.average(X_test_1[Y_test_1==0]**2,weights=w_test_1[Y_test_1==0]))\n",
    "print(np.mean(X_test_1[Y_test_1==1]),np.mean(X_test_1[Y_test_1==0]),np.average(X_test_1[Y_test_1==0],weights=w_test_1[Y_test_1==0]))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7ee1467f",
   "metadata": {},
   "source": [
    "## Now, try all moments at once"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 203,
   "id": "1f86619a",
   "metadata": {},
   "outputs": [],
   "source": [
    "from tensorflow.keras.layers import BatchNormalization"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 359,
   "id": "2a88f1e0",
   "metadata": {},
   "outputs": [],
   "source": [
    "#mymodel_inputtest_unifold = Input(shape=(1,))\n",
    "#hidden_layer_1_unifold = Dense(50, activation='LeakyReLU',use_bias=False)(mymodel_inputtest_unifold)\n",
    "#batch_norm_1_unifold = BatchNormalization()(hidden_layer_1_unifold)\n",
    "#hidden_layer_2_unifold = Dense(50, activation='LeakyReLU',use_bias=False)(batch_norm_1_unifold)\n",
    "#batch_norm_2_unifold = BatchNormalization()(hidden_layer_2_unifold)\n",
    "#hidden_layer_3_unifold = Dense(50, activation='LeakyReLU',use_bias=False)(batch_norm_2_unifold)\n",
    "#outputs_unifold = Dense(1, activation='sigmoid',use_bias=False)(hidden_layer_3_unifold) #tanh\n",
    "#model_generator_unifold = Model(mymodel_inputtest_unifold, outputs_unifold) #tf.exp(10*outputs_unifold))\n",
    "\n",
    "mymodel_inputtest_unifold = Input(shape=(1,))\n",
    "hidden_layer_1_unifold = Dense(50, activation='LeakyReLU',use_bias=True)(mymodel_inputtest_unifold)\n",
    "batch_norm_1_unifold = BatchNormalization()(hidden_layer_1_unifold)\n",
    "hidden_layer_2_unifold = Dense(50, activation='LeakyReLU',use_bias=True)(batch_norm_1_unifold)\n",
    "batch_norm_2_unifold = BatchNormalization()(hidden_layer_2_unifold)\n",
    "hidden_layer_3_unifold = Dense(50, activation='LeakyReLU',use_bias=True)(batch_norm_2_unifold)\n",
    "outputs_unifold = Dense(1, activation='sigmoid',use_bias=True)(hidden_layer_3_unifold) #tanh\n",
    "model_generator_unifold = Model(mymodel_inputtest_unifold, outputs_unifold)\n",
    "\n",
    "inputs_disc_unifold = Input((1, ))\n",
    "hidden_layer_1_disc_unifold = Dense(50, activation='LeakyReLU')(inputs_disc_unifold)\n",
    "batch_norm_1_disc_unifold = BatchNormalization()(hidden_layer_1_disc_unifold)\n",
    "hidden_layer_2_disc_unifold = Dense(50, activation='LeakyReLU')(batch_norm_1_disc_unifold)\n",
    "batch_norm_2_disc_unifold = BatchNormalization()(hidden_layer_2_disc_unifold)\n",
    "hidden_layer_3_disc_unifold = Dense(50, activation='LeakyReLU')(batch_norm_2_disc_unifold)\n",
    "outputs_disc_unifold = Dense(1, activation='sigmoid')(hidden_layer_3_disc_unifold)\n",
    "model_discrimantor_unifold = Model(inputs=inputs_disc_unifold, outputs=outputs_disc_unifold)\n",
    "\n",
    "opt = tf.keras.optimizers.Adam(learning_rate=1e-4)\n",
    "#opt2 = tf.keras.optimizers.Adam(learning_rate=1e-4)\n",
    "\n",
    "model_discrimantor_unifold.compile(loss=weighted_binary_crossentropy, optimizer=opt)\n",
    "    \n",
    "model_discrimantor_unifold.trainable = False\n",
    "mymodel_gan_unifold = Input(shape=(1,))\n",
    "gan_model_unifold = Model(inputs=mymodel_gan_unifold,outputs=concatenate([model_discrimantor_unifold(mymodel_gan_unifold),model_generator_unifold(mymodel_gan_unifold)]))\n",
    "\n",
    "gan_model_unifold.compile(loss=weighted_binary_crossentropy_GAN, optimizer=opt)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 360,
   "id": "911ec820",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "on epoch= 0 0.51998925 0.5000002 0.62410694\n",
      "on epoch= 1 0.5162633 0.28033668 0.5722849\n",
      "on epoch= 2 0.42468843 0.019470844 0.60550165\n",
      "on epoch= 3 0.33900538 0.0003531439 0.7093263\n",
      "on epoch= 4 0.41601765 1.2251841e-06 0.7488651\n",
      "on epoch= 5 0.2843844 0.054049026 1.0\n",
      "on epoch= 6 0.051905934 0.010275903 0.99999714\n",
      "on epoch= 7 0.5191289 4.293729e-18 1.0\n",
      "on epoch= 8 0.46522847 1.5785541e-05 0.9905676\n",
      "on epoch= 9 0.10908751 0.0028851437 0.9999945\n"
     ]
    }
   ],
   "source": [
    "xvals_1 = np.concatenate([gauss_data,gauss_sim])\n",
    "yvals_1 = np.concatenate([np.ones(len(gauss_data)),np.zeros(len(gauss_sim))])\n",
    "\n",
    "#xvals_1 = np.concatenate([w_true_alt,w_true])\n",
    "#yvals_1 = np.concatenate([np.ones(len(w_true_alt)),np.zeros(len(w_true))])\n",
    "\n",
    "X_std = (xvals_1 - xvals_1.min(axis=0)) / (xvals_1.max(axis=0) - xvals_1.min(axis=0))\n",
    "xvals_1 = X_std * (np.max(xvals_1) - np.min(xvals_1)) + np.min(xvals_1)\n",
    "\n",
    "X_train_1, X_test_1, Y_train_1, Y_test_1 = train_test_split(xvals_1, yvals_1)\n",
    "\n",
    "n_epochs = 10\n",
    "n_batch = 128*10\n",
    "n_batches = len(X_train_1) // n_batch\n",
    "\n",
    "for i in range(n_epochs):\n",
    "    mypreds = model_generator_unifold.predict(X_test_1,batch_size=1000)\n",
    "    print(\"on epoch=\",i,np.mean(mypreds),np.min(mypreds),np.max(mypreds))\n",
    "    for j in range(n_batches):\n",
    "        X_batch = X_train_1[j*n_batch:(j+1)*n_batch]\n",
    "        Y_batch = Y_train_1[j*n_batch:(j+1)*n_batch]\n",
    "        W_batch = model_generator_unifold(X_batch)\n",
    "        W_batch = np.array(W_batch).flatten()\n",
    "        W_batch[Y_batch==1] = 1\n",
    "        #W_batch[Y_batch==0] = 1\n",
    "        \n",
    "        mypreds = model_generator_unifold.predict(X_batch,batch_size=1000)\n",
    "        #print(\"     on batch=\",j,np.mean(mypreds),np.min(mypreds),np.max(mypreds))\n",
    "        \n",
    "        Y_batch_2 = np.stack((Y_batch, W_batch), axis=1)\n",
    "        \n",
    "        if (i%3==0):\n",
    "            model_discrimantor_unifold.train_on_batch(X_batch, Y_batch_2)\n",
    "        \n",
    "        gan_model_unifold.train_on_batch(X_batch[Y_batch==0],np.zeros(len(X_batch[Y_batch==0])))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 356,
   "id": "4ca90342",
   "metadata": {},
   "outputs": [],
   "source": [
    "mypreds = model_generator_unifold.predict(X_test_1,batch_size=1000)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 357,
   "id": "aa754ade",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5, 0, 'jet width')"
      ]
     },
     "execution_count": 357,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "#bins = np.linspace(0,1,30)\n",
    "bins = np.linspace(-3,3,20)\n",
    "_,_,_=plt.hist(X_test_1[Y_test_1==1],bins=bins,alpha=0.5,label=\"data\",density=True)\n",
    "_,_,_=plt.hist(X_test_1[Y_test_1==0],bins=bins,alpha=0.5,label=\"MC\",density=True)\n",
    "_,_,_=plt.hist(X_test_1[Y_test_1==0],bins=bins,weights=mypreds[Y_test_1==0],histtype=\"step\",color=\"black\",ls=\":\",label=\"weighted MC\",density=True)\n",
    "plt.legend(fontsize=15)\n",
    "plt.xlabel(\"jet width\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 358,
   "id": "dc38cc1e",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.collections.PathCollection at 0x7fc5440a7a00>"
      ]
     },
     "execution_count": 358,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.scatter(bins,model_generator_unifold.predict(bins))"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.13"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}