SSL_2DEmbedding_Clusters.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "white-branch",
   "metadata": {},
   "outputs": [],
   "source": [
    "import itertools\n",
    "from itertools import chain, combinations\n",
    "from WMMSE_SSL_Kshot import k_shot_SSL_train_test\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import matplotlib.cm as cm"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "super-monte",
   "metadata": {},
   "outputs": [],
   "source": [
    "# set the parameters\n",
    "n = 3 # number of TP-UE pairs\n",
    "num_train_samples = 10000 # total number of train samples\n",
    "num_test_samples = 1000 # number of test samples\n",
    "num_labeled_train_samples = 100 # number of labeled train samples (note that this parameter does not matter for the SSL pre-training phase)\n",
    "batch_size = 64 # train/test batch size\n",
    "hidden_layers = [128, 2] # backbone hidden layers\n",
    "num_SSL_pretrain_epochs = 20 # number of SSL pre-training epochs\n",
    "num_k_shot_epochs = 0 # number of supervised training epochs\n",
    "tau = 0.1 # temperatue in contrastive loss\n",
    "lr = 5e-2 # learning rate\n",
    "device = 'cuda:0' # the device (cpu/gpu) to perform computations on"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "preceding-tamil",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 20/20 [01:37<00:00,  4.87s/it]\n"
     ]
    }
   ],
   "source": [
    "# pre-train the backbone using SSL\n",
    "results = k_shot_SSL_train_test(n=n, num_train_samples=num_train_samples, num_test_samples=num_test_samples,\n",
    "                                num_labeled_train_samples=num_labeled_train_samples, batch_size=batch_size,\n",
    "                                hidden_layers=hidden_layers, num_SSL_pretrain_epochs=num_SSL_pretrain_epochs,\n",
    "                                num_k_shot_epochs=num_k_shot_epochs, tau=tau, lr=lr, device=device\n",
    "                               )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "offensive-huntington",
   "metadata": {},
   "outputs": [],
   "source": [
    "# retrieve the embeddings and ground-truth power levels of the training samples\n",
    "all_initial_embeddings = np.concatenate((results[2]['train-labeled'], results[2]['train-unlabeled']), axis=0)\n",
    "all_embeddings = np.concatenate((results[3]['train-labeled'], results[3]['train-unlabeled']), axis=0)\n",
    "all_groundtruth_PC_levels = np.concatenate((results[4]['train-labeled'], results[4]['train-unlabeled']), axis=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "confused-discovery",
   "metadata": {},
   "outputs": [],
   "source": [
    "# derive the binary power control labels\n",
    "def powerset(iterable):\n",
    "    \"powerset([1,2,3]) --> () (1,) (2,) (3,) (1,2) (1,3) (2,3) (1,2,3)\"\n",
    "    s = list(iterable)\n",
    "    return chain.from_iterable(combinations(s, r) for r in range(len(s)+1))\n",
    "\n",
    "cases = list(powerset(range(n)))[1:]\n",
    "ncol = n\n",
    "nrow = 2 ** n - 1\n",
    "all_binaries = np.zeros((nrow, ncol), int)\n",
    "all_binaries[np.arange(nrow).repeat([*map(len, cases)]), np.concatenate(cases)] = 1\n",
    "\n",
    "all_binary_PC_labels = []\n",
    "for y in all_groundtruth_PC_levels:\n",
    "    distances = np.linalg.norm(all_binaries - y, ord=2, axis=1)\n",
    "    all_binary_PC_labels.append(np.argmin(distances))\n",
    "all_binary_PC_labels = np.array(all_binary_PC_labels)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "musical-hopkins",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1080x360 with 14 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# plot the clusters before and after SSL pre-training\n",
    "colors = cm.rainbow(np.linspace(0, 1, 2 ** n - 1))\n",
    "plt_grid_size = int(np.ceil(np.sqrt(2 ** n - 1)))\n",
    "fig, ax = plt.subplots(2, 2 ** n - 1, figsize=(15, 5))\n",
    "\n",
    "for PC_label in range(2 ** n - 1):\n",
    "    ax_row = 0\n",
    "    ax_col = PC_label\n",
    "    ax[ax_row, ax_col].clear()\n",
    "    ax[ax_row, ax_col].set_title(str(all_binaries[PC_label]))\n",
    "    sample_indices = np.where(all_binary_PC_labels == PC_label)[0]\n",
    "    ax[ax_row, ax_col].scatter(all_initial_embeddings[sample_indices, 0],\n",
    "                               all_initial_embeddings[sample_indices, 1],\n",
    "                               s=0.01,\n",
    "                               color=colors[PC_label]\n",
    "                              )\n",
    "    ax[ax_row, ax_col].set_xlim([-1.1, 1.1])\n",
    "    ax[ax_row, ax_col].set_ylim([-1.1, 1.1])\n",
    "    ax[ax_row, ax_col].set_aspect('equal')\n",
    "    \n",
    "    if PC_label == 0:\n",
    "        ax[ax_row, ax_col].text(-5, -.1, 'Before training', fontsize=14, ha='center')\n",
    "    \n",
    "    ax_row = 1\n",
    "    ax_col = PC_label\n",
    "    ax[ax_row, ax_col].clear()\n",
    "    ax[ax_row, ax_col].set_title(str(all_binaries[PC_label]))\n",
    "    sample_indices = np.where(all_binary_PC_labels == PC_label)[0]\n",
    "    ax[ax_row, ax_col].scatter(all_embeddings[sample_indices, 0],\n",
    "                               all_embeddings[sample_indices, 1],\n",
    "                               s=0.01,\n",
    "                               color=colors[PC_label]\n",
    "                              )\n",
    "    ax[ax_row, ax_col].set_xlim([-1.1, 1.1])\n",
    "    ax[ax_row, ax_col].set_ylim([-1.1, 1.1])\n",
    "    ax[ax_row, ax_col].set_aspect('equal')\n",
    "    if PC_label == 0:\n",
    "        ax[ax_row, ax_col].text(-5, -.1, 'After self-supervised pre-training', fontsize=14, ha='center')\n",
    "        \n",
    "plt.subplots_adjust(left=None, bottom=None, right=None, top=None, wspace=.5, hspace=None)\n",
    "plt.savefig('SSL_clusters.pdf', bbox_inches='tight')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "identified-firewall",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.7.9"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}