线性回归的tensorflow实现.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "heading_collapsed": true
   },
   "source": [
    "# 线性回归"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "code_folding": [],
    "hidden": true,
    "run_control": {
     "marked": false
    }
   },
   "outputs": [],
   "source": [
    "#数据集：Iris\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "import tensorflow as tf\n",
    "from sklearn import datasets\n",
    "from tensorflow.python.framework import ops\n",
    "\n",
    "ops.reset_default_graph()\n",
    "sess = tf.Session()\n",
    "iris = datasets.load_iris()\n",
    "\n",
    "x_vals = np.array([x[3] for x in iris.data])\n",
    "y_vals = np.array([y[0] for y in iris.data])\n",
    "\n",
    "learning_rate = 0.05\n",
    "batch_size = 25\n",
    "x_data = tf.placeholder(tf.float32, shape=[None, 1])\n",
    "y_data = tf.placeholder(tf.float32, shape=[None, 1])\n",
    "A = tf.Variable(tf.random_normal(shape=[1, 1]))\n",
    "b = tf.Variable(tf.random_normal(shape=[1, 1]))\n",
    "model_output = tf.matmul(x_data, A) + b\n",
    "#L2 loss\n",
    "loss_L2 = tf.reduce_mean(tf.square(y_data - model_output))\n",
    "#L1 loss\n",
    "#loss = tf.reduce_mean(tf.abs(y_data - model_output))\n",
    "sess.run(tf.global_variables_initializer())\n",
    "train_step = tf.train.GradientDescentOptimizer(learning_rate).minimize(loss_L2)\n",
    "\n",
    "loss_L2_vec = []\n",
    "A_vec = []\n",
    "b_vec = []\n",
    "for i in range(100):\n",
    "    rand_index = np.random.choice(len(x_vals), size=batch_size, replace=False)\n",
    "    rand_x = np.transpose([x_vals[rand_index]])\n",
    "    rand_y = np.transpose([y_vals[rand_index]])\n",
    "    sess.run(train_step, feed_dict={x_data: rand_x, y_data: rand_y})\n",
    "    loss_temp = sess.run(loss_L2, feed_dict={x_data: rand_x, y_data: rand_y})\n",
    "    loss_L2_vec.append(loss_temp)\n",
    "    A_temp = sess.run(A)\n",
    "    A_vec.append(A_temp)\n",
    "    b_temp = sess.run(b)\n",
    "    b_vec.append(b_temp)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "hidden": true,
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x2a3911a6898>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(loss_L2_vec)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "code_folding": [],
    "hidden": true
   },
   "outputs": [],
   "source": [
    "#数据集：Iris\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "import tensorflow as tf\n",
    "from sklearn import datasets\n",
    "from tensorflow.python.framework import ops\n",
    "\n",
    "ops.reset_default_graph()\n",
    "sess = tf.Session()\n",
    "iris = datasets.load_iris()\n",
    "\n",
    "x_vals = np.array([x[3] for x in iris.data])\n",
    "y_vals = np.array([y[0] for y in iris.data])\n",
    "\n",
    "learning_rate = 0.05\n",
    "batch_size = 25\n",
    "x_data = tf.placeholder(tf.float32, shape=[None, 1])\n",
    "y_data = tf.placeholder(tf.float32, shape=[None, 1])\n",
    "A = tf.Variable(tf.random_normal(shape=[1, 1]))\n",
    "b = tf.Variable(tf.random_normal(shape=[1, 1]))\n",
    "model_output = tf.matmul(x_data, A) + b\n",
    "#L1 loss\n",
    "loss_L1 = tf.reduce_mean(tf.abs(y_data - model_output))\n",
    "sess.run(tf.global_variables_initializer())\n",
    "train_step = tf.train.GradientDescentOptimizer(learning_rate).minimize(loss_L1)\n",
    "\n",
    "loss_L1_vec = []\n",
    "A_vec = []\n",
    "b_vec = []\n",
    "for i in range(100):\n",
    "    rand_index = np.random.choice(len(x_vals), size=batch_size, replace=False)\n",
    "    rand_x = np.transpose([x_vals[rand_index]])\n",
    "    rand_y = np.transpose([y_vals[rand_index]])\n",
    "    sess.run(train_step, feed_dict={x_data: rand_x, y_data: rand_y})\n",
    "    loss_temp = sess.run(loss_L1, feed_dict={x_data: rand_x, y_data: rand_y})\n",
    "    loss_L1_vec.append(loss_temp)\n",
    "    A_temp = sess.run(A)\n",
    "    A_vec.append(A_temp)\n",
    "    b_temp = sess.run(b)\n",
    "    b_vec.append(b_temp)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "hidden": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x2a38e6086a0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(loss_L1_vec)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "hidden": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x2a3914dd1d0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(loss_L1_vec, 'r--', label='L1_loss')\n",
    "plt.plot(loss_L2_vec, 'b-', label='L2_loss')\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "heading_collapsed": true
   },
   "source": [
    "# Lasso"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "hidden": true
   },
   "outputs": [],
   "source": [
    "#数据集：Iris\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "import tensorflow as tf\n",
    "from sklearn import datasets\n",
    "from tensorflow.python.framework import ops\n",
    "\n",
    "ops.reset_default_graph()\n",
    "sess = tf.Session()\n",
    "iris = datasets.load_iris()\n",
    "\n",
    "x_vals = np.array([x[3] for x in iris.data])\n",
    "y_vals = np.array([y[0] for y in iris.data])\n",
    "\n",
    "learning_rate = 0.001\n",
    "batch_size = 50\n",
    "x_data = tf.placeholder(tf.float32, shape=[None, 1])\n",
    "y_data = tf.placeholder(tf.float32, shape=[None, 1])\n",
    "A = tf.Variable(tf.random_normal(shape=[1, 1]))\n",
    "b = tf.Variable(tf.random_normal(shape=[1, 1]))\n",
    "model_output = tf.matmul(x_data, A) + b\n",
    "#Loss_Lasso\n",
    "lasso_param = tf.constant(0.9)\n",
    "heavyside_step = tf.truediv(1.,\n",
    "                            tf.add(1.,\n",
    "                                   tf.exp(\n",
    "                                       tf.multiply(-100.,\n",
    "                                                   tf.subtract(A,\n",
    "                                                               lasso_param)))))\n",
    "regularization_parm = tf.multiply(heavyside_step, 99.)\n",
    "loss_lasso = tf.add(\n",
    "    tf.reduce_mean(tf.square(y_data - model_output)), regularization_parm)\n",
    "\n",
    "sess.run(tf.global_variables_initializer())\n",
    "train_step = tf.train.GradientDescentOptimizer(learning_rate).minimize(\n",
    "    loss_lasso)\n",
    "\n",
    "loss_lasso_vec = []\n",
    "A_vec = []\n",
    "b_vec = []\n",
    "for i in range(1500):\n",
    "    rand_index = np.random.choice(len(x_vals), size=batch_size, replace=False)\n",
    "    rand_x = np.transpose([x_vals[rand_index]])\n",
    "    rand_y = np.transpose([y_vals[rand_index]])\n",
    "    sess.run(train_step, feed_dict={x_data: rand_x, y_data: rand_y})\n",
    "    loss_temp = sess.run(\n",
    "        loss_lasso, feed_dict={\n",
    "            x_data: rand_x,\n",
    "            y_data: rand_y\n",
    "        })\n",
    "    loss_temp1 = loss_temp[0][0]\n",
    "    loss_lasso_vec.append(loss_temp1)\n",
    "    A_temp = sess.run(A)\n",
    "    A_vec.append(A_temp)\n",
    "    b_temp = sess.run(b)\n",
    "    b_vec.append(b_temp)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "hidden": true,
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x2a39111eb00>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(loss_lasso_vec, 'b--', label = 'Lasso_loss')\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "hidden": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "heading_collapsed": true
   },
   "source": [
    "# 岭回归"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "hidden": true
   },
   "outputs": [],
   "source": [
    "#数据集：Iris\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "import tensorflow as tf\n",
    "from sklearn import datasets\n",
    "from tensorflow.python.framework import ops\n",
    "\n",
    "ops.reset_default_graph()\n",
    "sess = tf.Session()\n",
    "iris = datasets.load_iris()\n",
    "\n",
    "x_vals = np.array([x[3] for x in iris.data])\n",
    "y_vals = np.array([y[0] for y in iris.data])\n",
    "\n",
    "learning_rate = 0.001\n",
    "batch_size = 50\n",
    "x_data = tf.placeholder(tf.float32, shape=[None, 1])\n",
    "y_data = tf.placeholder(tf.float32, shape=[None, 1])\n",
    "A = tf.Variable(tf.random_normal(shape=[1, 1]))\n",
    "b = tf.Variable(tf.random_normal(shape=[1, 1]))\n",
    "model_output = tf.matmul(x_data, A) + b\n",
    "#Ridge_loss\n",
    "ridge_param = tf.constant(0.9)\n",
    "loss_penalty = tf.reduce_mean(tf.square(A))\n",
    "loss_ridge = tf.expand_dims(\n",
    "    tf.add(\n",
    "        tf.reduce_mean(tf.square(y_data - model_output)),\n",
    "        tf.multiply(ridge_param, loss_penalty)), 0)\n",
    "\n",
    "sess.run(tf.global_variables_initializer())\n",
    "train_step = tf.train.GradientDescentOptimizer(learning_rate).minimize(\n",
    "    loss_ridge)\n",
    "\n",
    "loss_ridge_vec = []\n",
    "A_vec = []\n",
    "b_vec = []\n",
    "for i in range(1500):\n",
    "    rand_index = np.random.choice(len(x_vals), size=batch_size, replace=False)\n",
    "    rand_x = np.transpose([x_vals[rand_index]])\n",
    "    rand_y = np.transpose([y_vals[rand_index]])\n",
    "    sess.run(train_step, feed_dict={x_data: rand_x, y_data: rand_y})\n",
    "    loss_temp = sess.run(\n",
    "        loss_ridge, feed_dict={\n",
    "            x_data: rand_x,\n",
    "            y_data: rand_y\n",
    "        })\n",
    "    #loss_temp1 = loss_temp[0][0]\n",
    "    loss_ridge_vec.append(loss_temp1)\n",
    "    A_temp = sess.run(A)\n",
    "    A_vec.append(A_temp)\n",
    "    b_temp = sess.run(b)\n",
    "    b_vec.append(b_temp)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "hidden": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x2a3915c8438>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(loss_ridge_vec, 'r-', label = 'Ridge_loss')\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "hidden": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x2a3915de8d0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(loss_ridge_vec, 'r-', label = 'Ridge_loss')\n",
    "plt.plot(loss_lasso_vec, 'b--', label = 'Lasso_loss')\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "heading_collapsed": true
   },
   "source": [
    "# 4tensorboard"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "hidden": true
   },
   "outputs": [
    {
     "ename": "ImportError",
     "evalue": "Traceback (most recent call last):\n  File \"F:\\Anaconda3-5-1\\envs\\tensorflow\\lib\\site-packages\\tensorflow\\python\\pywrap_tensorflow.py\", line 58, in <module>\n    from tensorflow.python.pywrap_tensorflow_internal import *\n  File \"F:\\Anaconda3-5-1\\envs\\tensorflow\\lib\\site-packages\\tensorflow\\python\\pywrap_tensorflow_internal.py\", line 21, in <module>\n    _pywrap_tensorflow_internal = swig_import_helper()\n  File \"F:\\Anaconda3-5-1\\envs\\tensorflow\\lib\\site-packages\\tensorflow\\python\\pywrap_tensorflow_internal.py\", line 20, in swig_import_helper\n    return importlib.import_module('_pywrap_tensorflow_internal')\n  File \"F:\\Anaconda3-5-1\\envs\\tensorflow\\lib\\importlib\\__init__.py\", line 37, in import_module\n    __import__(name)\nImportError: No module named _pywrap_tensorflow_internal\n\n\nFailed to load the native TensorFlow runtime.\n\nSee https://www.tensorflow.org/install/install_sources#common_installation_problems\n\nfor some common reasons and solutions.  Include the entire stack trace\nabove this error message when asking for help.",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mImportError\u001b[0m                               Traceback (most recent call last)",
      "\u001b[1;32m<ipython-input-3-d6627d7d3743>\u001b[0m in \u001b[0;36m<module>\u001b[1;34m()\u001b[0m\n\u001b[0;32m      2\u001b[0m \u001b[1;32mimport\u001b[0m \u001b[0mmatplotlib\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mpyplot\u001b[0m \u001b[1;32mas\u001b[0m \u001b[0mplt\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m      3\u001b[0m \u001b[1;32mimport\u001b[0m \u001b[0mnumpy\u001b[0m \u001b[1;32mas\u001b[0m \u001b[0mnp\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m----> 4\u001b[1;33m \u001b[1;32mimport\u001b[0m \u001b[0mtensorflow\u001b[0m \u001b[1;32mas\u001b[0m \u001b[0mtf\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m      5\u001b[0m \u001b[1;32mfrom\u001b[0m \u001b[0msklearn\u001b[0m \u001b[1;32mimport\u001b[0m \u001b[0mdatasets\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m      6\u001b[0m \u001b[1;32mfrom\u001b[0m \u001b[0mtensorflow\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mpython\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mframework\u001b[0m \u001b[1;32mimport\u001b[0m \u001b[0mops\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mF:\\Anaconda3-5-1\\envs\\tensorflow\\lib\\site-packages\\tensorflow\\__init__.py\u001b[0m in \u001b[0;36m<module>\u001b[1;34m()\u001b[0m\n\u001b[0;32m     22\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     23\u001b[0m \u001b[1;31m# pylint: disable=wildcard-import\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m---> 24\u001b[1;33m \u001b[1;32mfrom\u001b[0m \u001b[0mtensorflow\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mpython\u001b[0m \u001b[1;32mimport\u001b[0m \u001b[1;33m*\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m     25\u001b[0m \u001b[1;31m# pylint: enable=wildcard-import\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     26\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mF:\\Anaconda3-5-1\\envs\\tensorflow\\lib\\site-packages\\tensorflow\\python\\__init__.py\u001b[0m in \u001b[0;36m<module>\u001b[1;34m()\u001b[0m\n\u001b[0;32m     47\u001b[0m \u001b[1;32mimport\u001b[0m \u001b[0mnumpy\u001b[0m \u001b[1;32mas\u001b[0m \u001b[0mnp\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     48\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m---> 49\u001b[1;33m \u001b[1;32mfrom\u001b[0m \u001b[0mtensorflow\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mpython\u001b[0m \u001b[1;32mimport\u001b[0m \u001b[0mpywrap_tensorflow\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m     50\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     51\u001b[0m \u001b[1;31m# Protocol buffers\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32mF:\\Anaconda3-5-1\\envs\\tensorflow\\lib\\site-packages\\tensorflow\\python\\pywrap_tensorflow.py\u001b[0m in \u001b[0;36m<module>\u001b[1;34m()\u001b[0m\n\u001b[0;32m     72\u001b[0m \u001b[1;32mfor\u001b[0m \u001b[0msome\u001b[0m \u001b[0mcommon\u001b[0m \u001b[0mreasons\u001b[0m \u001b[1;32mand\u001b[0m \u001b[0msolutions\u001b[0m\u001b[1;33m.\u001b[0m  \u001b[0mInclude\u001b[0m \u001b[0mthe\u001b[0m \u001b[0mentire\u001b[0m \u001b[0mstack\u001b[0m \u001b[0mtrace\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     73\u001b[0m above this error message when asking for help.\"\"\" % traceback.format_exc()\n\u001b[1;32m---> 74\u001b[1;33m   \u001b[1;32mraise\u001b[0m \u001b[0mImportError\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mmsg\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m     75\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     76\u001b[0m \u001b[1;31m# pylint: enable=wildcard-import,g-import-not-at-top,unused-import,line-too-long\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;31mImportError\u001b[0m: Traceback (most recent call last):\n  File \"F:\\Anaconda3-5-1\\envs\\tensorflow\\lib\\site-packages\\tensorflow\\python\\pywrap_tensorflow.py\", line 58, in <module>\n    from tensorflow.python.pywrap_tensorflow_internal import *\n  File \"F:\\Anaconda3-5-1\\envs\\tensorflow\\lib\\site-packages\\tensorflow\\python\\pywrap_tensorflow_internal.py\", line 21, in <module>\n    _pywrap_tensorflow_internal = swig_import_helper()\n  File \"F:\\Anaconda3-5-1\\envs\\tensorflow\\lib\\site-packages\\tensorflow\\python\\pywrap_tensorflow_internal.py\", line 20, in swig_import_helper\n    return importlib.import_module('_pywrap_tensorflow_internal')\n  File \"F:\\Anaconda3-5-1\\envs\\tensorflow\\lib\\importlib\\__init__.py\", line 37, in import_module\n    __import__(name)\nImportError: No module named _pywrap_tensorflow_internal\n\n\nFailed to load the native TensorFlow runtime.\n\nSee https://www.tensorflow.org/install/install_sources#common_installation_problems\n\nfor some common reasons and solutions.  Include the entire stack trace\nabove this error message when asking for help."
     ]
    }
   ],
   "source": [
    "#数据集：Iris\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "import tensorflow as tf\n",
    "from sklearn import datasets\n",
    "from tensorflow.python.framework import ops\n",
    "\n",
    "ops.reset_default_graph()\n",
    "\n",
    "\n",
    "\n",
    "iris = datasets.load_iris()\n",
    "x_vals = np.array([x[3] for x in iris.data])\n",
    "y_vals = np.array([y[0] for y in iris.data])\n",
    "\n",
    "learning_rate = 0.05\n",
    "batch_size = 25\n",
    "\n",
    "with tf.name_scope('data'):\n",
    "    x_data = tf.placeholder(tf.float32, shape=[None, 1])\n",
    "    y_data = tf.placeholder(tf.float32, shape=[None, 1])\n",
    "    \n",
    "with tf.name_scope('parameters'):\n",
    "    A = tf.Variable(tf.random_normal(shape=[1, 1]))\n",
    "    b = tf.Variable(tf.random_normal(shape=[1, 1]))\n",
    "\n",
    "with tf.name_scope('prediction'):\n",
    "    model_output = tf.matmul(x_data, A) + b\n",
    "#L2 loss\n",
    "with tf.name_scope('loss'):\n",
    "    loss_L2 = tf.reduce_mean(tf.square(y_data - model_output))\n",
    "#L1 loss\n",
    "#loss = tf.reduce_mean(tf.abs(y_data - model_output))\n",
    "with tf.name_scope('init'):\n",
    "    init = tf.global_variables_initializer()\n",
    "\n",
    "    \n",
    "optimizer = tf.train.GradientDescentOptimizer(learning_rate)\n",
    "with tf.name_scope('train'):\n",
    "    train_step = optimizer.minimize(loss_L2)\n",
    "\n",
    "sess = tf.Session()  \n",
    "writer = tf.summary.FileWriter(\"D:/My_python_Data/Jupyter/logs/\", sess.graph)\n",
    "sess.run(init)\n",
    "loss_L2_vec = []\n",
    "A_vec = []\n",
    "b_vec = []\n",
    "for i in range(100):\n",
    "    rand_index = np.random.choice(len(x_vals), size=batch_size, replace=False)\n",
    "    rand_x = np.transpose([x_vals[rand_index]])\n",
    "    rand_y = np.transpose([y_vals[rand_index]])\n",
    "    sess.run(train_step, feed_dict={x_data: rand_x, y_data: rand_y})\n",
    "    \n",
    "\n",
    "    loss_temp = sess.run(loss_L2, feed_dict={x_data: rand_x, y_data: rand_y})\n",
    "    loss_L2_vec.append(loss_temp)\n",
    "    A_temp = sess.run(A)\n",
    "    A_vec.append(A_temp)\n",
    "    b_temp = sess.run(b)\n",
    "    b_vec.append(b_temp)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "hidden": true
   },
   "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": "3.6.4"
  },
  "toc": {
   "base_numbering": 1,
   "nav_menu": {},
   "number_sections": true,
   "sideBar": true,
   "skip_h1_title": false,
   "title_cell": "Table of Contents",
   "title_sidebar": "Contents",
   "toc_cell": false,
   "toc_position": {},
   "toc_section_display": true,
   "toc_window_display": true
  },
  "varInspector": {
   "cols": {
    "lenName": 16,
    "lenType": 16,
    "lenVar": 40
   },
   "kernels_config": {
    "python": {
     "delete_cmd_postfix": "",
     "delete_cmd_prefix": "del ",
     "library": "var_list.py",
     "varRefreshCmd": "print(var_dic_list())"
    },
    "r": {
     "delete_cmd_postfix": ") ",
     "delete_cmd_prefix": "rm(",
     "library": "var_list.r",
     "varRefreshCmd": "cat(var_dic_list()) "
    }
   },
   "types_to_exclude": [
    "module",
    "function",
    "builtin_function_or_method",
    "instance",
    "_Feature"
   ],
   "window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}