visualize_nn.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "import tensorflow as tf\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "from matplotlib.animation import FuncAnimation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "x_min = y_min = -2\n",
    "x_max = y_max = 2\n",
    "learning_rate = 0.05\n",
    "num_epochs = 10000"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "def plot1d(d1, d2):\n",
    "    min_x = min(min(d1),min(d2))\n",
    "    max_x = max(max(d1),max(d2))\n",
    "    bound_x = (max_x - min_x) / 5\n",
    "    plt.xlim(min_x-bound_x, max_x+bound_x)\n",
    "    plt.ylim(min_x-bound_x, max_x+bound_x)\n",
    "    plt.scatter(d1, [0]*len(d1), c='b', marker='.', s=5)\n",
    "    plt.scatter(d2, [0]*len(d2), c='r', marker='.', s=5)\n",
    "    plt.show()\n",
    "    \n",
    "def plot2d(d1, d2):\n",
    "    min_x = min(min(d1[0]),min(d2[0]))\n",
    "    max_x = max(max(d1[0]),max(d2[0]))\n",
    "    bound_x = (max_x - min_x) / 5\n",
    "    min_y = min(min(d1[1]),min(d2[1]))\n",
    "    max_y = max(max(d1[1]),max(d2[1]))\n",
    "    bound_y = (max_y - min_y) / 5\n",
    "    plt.xlim(min_x-bound_x, max_x+bound_x)\n",
    "    plt.ylim(min_y-bound_y, max_y+bound_y)\n",
    "    plt.scatter(d1[0], d1[1], c='b', marker='.', s=5)\n",
    "    plt.scatter(d2[0], d2[1], c='r', marker='.', s=5)\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "x_a = np.mgrid[-1:1:0.01].reshape(-1)\n",
    "x_b = np.concatenate((np.mgrid[-3:-2:0.01].reshape(-1), np.mgrid[2:3:0.01].reshape(-1)), axis=0)\n",
    "x_labels = np.concatenate((x_a, x_b), axis=0)\n",
    "np.random.shuffle(x_labels)\n",
    "y_labels = np.asarray(map(lambda x: [1,0] if x < 1 and x >= -1 else [0,1], x_labels))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "def model(X):\n",
    "    X_1 = tf.expand_dims(X, 0)\n",
    "    W_1 = tf.Variable(tf.random_normal([2,1])) # [[3.0],[1.0]]\n",
    "    B_1 = tf.Variable(tf.random_normal([2,1]))\n",
    "    L_1 = tf.add(tf.matmul(W_1, X_1), B_1)\n",
    "    X_2 = tf.sigmoid(L_1)\n",
    "    W_2 = tf.Variable(tf.random_normal([2,2])) # [[3.0],[1.0]]\n",
    "    B_2 = tf.Variable(tf.random_normal([2,1]))\n",
    "    L_2 = tf.add(tf.matmul(W_2, X_2), B_2)\n",
    "    Y = tf.transpose(L_2)\n",
    "    return X_1, L_1, X_2, L_2, Y"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "def predict(Y):\n",
    "    a = Y[0]\n",
    "    b = Y[1]\n",
    "    if a > b:\n",
    "        return 'blue'\n",
    "    else:\n",
    "        return 'red'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "def train(graph, sess):\n",
    "    with graph.as_default():\n",
    "        X = tf.placeholder('float')\n",
    "        Y_labels = tf.placeholder('float')\n",
    "        X_1, L_1, X_2, L_2, Y = model(X)\n",
    "        cost = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels=Y_labels, logits=Y))\n",
    "        optimizer = tf.train.GradientDescentOptimizer(learning_rate).minimize(cost)\n",
    "        with sess.as_default():\n",
    "            sess.run(tf.global_variables_initializer())\n",
    "            feed_dict = {X: x_labels, Y_labels: y_labels}\n",
    "            for i in range(num_epochs):\n",
    "                _, loss = sess.run([optimizer, cost], feed_dict=feed_dict)\n",
    "                if i % 1000 == 0:\n",
    "                    print 'Loss: {}'.format(loss)\n",
    "                    y_a = sess.run(Y, feed_dict={X: x_a})\n",
    "                    y_b = sess.run(Y, feed_dict={X: x_b})\n",
    "                    plot2d(y_a.T, y_b.T)\n",
    "            return X, L_1, X_2, L_2, Y"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.905274808407\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f1b7af2ea10>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.619085609913\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f1b79ccc0d0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.533021569252\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f1b79d62d10>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.308309882879\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f1bd010a310>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0987442210317\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYAAAAD8CAYAAAB+UHOxAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMS4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvNQv5yAAAGEVJREFUeJzt3X+QVeWd5/H3R0DHjEaJ3SrgJWiFkJXUpnC6bPxRjInORjQFJiaW+UPMFBZr0NpYNTtZjVXZrVRl1iRb2VRWNMOA5Y9KJWYVQs8ER+Ook5mJ9NLyQ2gIDqHUBom0YBDiKCF894/n3OHS9O1f9/S9t/t8XlW3PPfcp8/zvZf2fvp5nnPPVURgZmbFc0qjCzAzs8ZwAJiZFZQDwMysoBwAZmYF5QAwMysoB4CZWUE5AMzMCsoBYGZWUA4AM7OCmtjoAgbS0tISM2bMaHQZZmZjxksvvfRWRLQOpW1TB8CMGTPo6upqdBlmZmOGpNeG2tZTQGZmBeUAMDMrKAeAmVlBOQDMzAqq5gCQVJL0vKRtkrolfaWfNldJOihpU3b7eq39mplZbfI4C+go8BcRsUHSmcBLkn4eEdv6tPuniPhMDv2ZmVkOah4BRMTeiNiQbR8CtgPTaj2umZmNrlzXACTNAOYAnf08fJmkzZKekjQ7z37NzGz4cvsgmKQzgCeBuyLinT4PbwA+HBGHJV0H/BSYWeU4S4AlANOnT8+rPDMz6yOXEYCkSaQ3/x9GxKq+j0fEOxFxONteC0yS1NLfsSJieUS0RURba+uQPs1sZmYjkMdZQAJWAtsj4rtV2pyftUPSpVm/+2vt28zMRi6PKaArgFuALZI2Zfu+BkwHiIgfAJ8HvizpKPBvwM0RETn0bWZmI1RzAETEPwMapM39wP219mVmZvnxJ4HNzArKAWBmVlAOADOzgnIAmJkVlAPAzKygHABmZgXlADAzKygHgJlZQTkAzMwKygFgZlZQDgAzs4JyAJiZFZQDwMysoBwAZmYF5QAwMysoB8BY0tkJc+em/5qZ1cgBMJZ85jPHQ+D66xtdjZmNcXl8J3BJ0vOStknqlvSVftpI0vcl7ZT0sqRLau23kN566/j22rVwzjkeDZjZiOUxAjgK/EVEXAzMBe6QdHGfNvOBmdltCfBgDv0Wz623nnj/wIE0Gli5sjH1mNmYVnMARMTeiNiQbR8CtgPT+jRbCDwayTrgbElTau27cB5+GNatg5aWE/ffdht85zsNKcnMxq5c1wAkzQDmAH3nJaYBPRX3d3NySJSPsURSl6Su3t7ePMsbH9rbobcXvv3tE/d/9atw4YWeEjKzIcstACSdATwJ3BUR74z0OBGxPCLaIqKttbU1r/LGn7/8S3j9dfjkJ4/ve/VVuOIK6OhoWFlmNnbkEgCSJpHe/H8YEav6abIHKFXcvyDbZ7UoleC552DFiuP7/vAHuOEGuPFG6Omp/rNmVnh5nAUkYCWwPSK+W6VZB7AoOxtoLnAwIvbW2rdlFi9OawOzZqX7EbBqFcyZ4ykhM6sqjxHAFcAtwKckbcpu10m6XdLtWZu1wC5gJ/A3wNIc+rVK7e3wq1/BmjUwcWLat38/XHmlRwNm1i9FRKNrqKqtrS26uroaXcbY09mZzgzatg2OHUv7zj03rQ20tze2NjMbVZJeioi2obT1J4HHo/Z22LIFVq+GD3wg7du3Dz71KVi0yKMBMwMcAOPbggVpWmjRIjj9dHj3XXjsMVi40CFgZg6Aca9Ugkcegeefh/POS/s2boS2Ni8QmxWcA6Ao2tth/fp0ZhCkKaGrr3YImBWYA6BISqV0llB5Suh3v0vTQUuXekrIrIAcAEVTOSU0dSq8+SY8+CDMn+8gMCsYB0BRtbenD4+Vp4S6u1MQeIHYrDAcAEVWnhJauhRmz077Nm6Em25yCJgVgAOg6EolWLYMnnoqBcGcOWlk0N7uBWKzcc4BYEk5CNasgSlTYO9eLxCbjXMOADtRqZQ+QVy5QOwgMBuXHAB2svICcXlKaOPGFAReGzAbVxwA1r/KKaHyIvG6dT5LyGwccQDYwMpBMG9eur9xIzzwQGNrMrNcTGx0ATZG3HMPSGl7qb/OwWw8cADY0JRHAmY2bngKyMysoPL6UviHJO2TtLXK41dJOljxlZFfz6NfMzMbubymgB4G7gceHaDNP0XEZ3Lqz8zMapTLCCAifgEcyONYZmZWH/VcA7hM0mZJT0maXa2RpCWSuiR19fb21rE8M7NiqVcAbAA+HBGfAP4P8NNqDSNieUS0RURba2trncozMyueugRARLwTEYez7bXAJEkt9ejbzMz6V5cAkHS+lD5FJOnSrN/99ejbzMz6l8tZQJJ+BFwFtEjaDfx3YBJARPwA+DzwZUlHgX8Dbo6IyKNvMzMbmVwCICK+OMjj95NOEzUzsybhTwKbmRWUA8DMrKAcAGZmBeUAMDMrKAeAmVlBOQDMzArKAWBmVlAOADOzgnIAmJkVlAPAzKygHABmZgXlADAzKygHgJlZQTkAzMwKygFgZlZQDoDR0NMDS5emW09Po6sxM+tXLgEg6SFJ+yRtrfK4JH1f0k5JL0u6JI9+m9ayZfDgg+k2fTpMmwadnY2uyszsBHmNAB4Grh3g8fnAzOy2BHgwp36b0x13nHj/jTdg7lz44Aeho6MxNZmZ9ZFLAETEL4ADAzRZCDwayTrgbElT8ui7KZVK8LWvnbz/0CFYuBBOOw1Wrqx/XWZmFeq1BjANqJwM353tG7+++U2IgDVr4MwzT3zsyBG47bY0IrjxRq8TmFlDNN0isKQlkrokdfX29ja6nNotWADvvAOvvw7XXguTJh1/7NAhWLUKZs50EJhZ3dUrAPYApYr7F2T7ThIRyyOiLSLaWltb61JcXZRK8NRT6a//FSvSNJCUHnv//RQEH/kIzJrlBWMzq4t6BUAHsCg7G2gucDAi9tap7+azeDG89x68+CJ89KNwSvbPcOQIvPIKXH65g8DMRl1ep4H+CHgRmCVpt6TFkm6XdHvWZC2wC9gJ/A2wNI9+x7z2dtixA375yxQEZ50FEybAsWMpCP70T1MQLFrk6SEzy50iotE1VNXW1hZdXV2NLqO+OjrSCOHw4TRKKJs8GaZOTWcPtbc3rj4za2qSXoqItqG0bbpF4MJbsAB6e+GFF2D27PTGD/D229DdnUYFH/+4p4fMrGYOgGbV3g5bt8LmzXDLLWmK6PTT04Kxg8DMcuAAaHalEjz6aForeP75NCqoDILrr0/7vE5gZsPkABhLyqOCchC0tMD+/bBtGzz2GMyf7wvQmdmQOQDGonIQbNiQpodaWtL+7u50AbprroE5czw9ZGYDcgCMZeXpoQ0b0l/+s2en/a+8Aps2pekhB4GZVTGx0QVYDkqldAnqnh647z74zW/gX/4F3nwzTREtWACf/jSccQbcc09qb2aF588BjFc9PXDvvfDMMykIymbPhnnzHARm45Q/B2DHp4fWrz9xeqi8TvDAA42tz8wazlNA413f6aFDh9LlqZf6ahxmRecAKIpyEJiZZTwFZGZWUA4AM7OCcgCYmRWUA8DMrKAcAGZmBeUAMDMrqLy+EvJaSTsk7ZR0dz+Pf0lSr6RN2e22PPo1M7ORq/lzAJImAMuAPwN2A+sldUTEtj5NH4+IO2vtz8zM8pHHCOBSYGdE7IqII8CPgYU5HNfMzEZRHgEwDaj8BpLd2b6+bpT0sqQnJPkqZGZmDVavReC/BWZExH8Efg48Uq2hpCWSuiR19fb21qk8M7PiySMA9gCVf9FfkO37dxGxPyLez+6uAP6k2sEiYnlEtEVEW2traw7lmZlZf/IIgPXATEkXSjoVuBnoqGwgaUrF3QXA9hz6NTOzGtR8FlBEHJV0J/A0MAF4KCK6JX0D6IqIDuC/SFoAHAUOAF+qtV8zM6uNvxHMzGwc8TeCmZnZoBwAZmYF5QAwMysoB4CZWUE5AMzMCsoBYGZWUA4AM7OCcgCYmRWUA8DMrKAcAGZmBeUAMDMrKAeAmVlBOQDMzArKAWBmVlAOADOzgnIAmJkVlAPAzKygcgkASddK2iFpp6S7+3n8NEmPZ493SpqRR79mZjZyNQeApAnAMmA+cDHwRUkX92m2GHg7Ij4C/G/gW7X2a2ZmtcljBHApsDMidkXEEeDHwMI+bRYCj2TbTwBXS1IOfZuZ2QjlEQDTgJ6K+7uzff22iYijwEHgnBz6NjOzEWq6RWBJSyR1Serq7e1tdDlmZuNWHgGwByhV3L8g29dvG0kTgbOA/f0dLCKWR0RbRLS1trbmUJ6ZmfUnjwBYD8yUdKGkU4GbgY4+bTqAW7PtzwPPRUTk0LeZmY3QxFoPEBFHJd0JPA1MAB6KiG5J3wC6IqIDWAk8JmkncIAUEmZm1kA1BwBARKwF1vbZ9/WK7feAL+TRl5mZ5aPpFoHNzKw+HABmZgXlADAzKygHgJlZQTkAzMwKygFgZlZQDgAzs4JyAJiZFZQDwGyc6OmBpUvTradn8PZmuXwS2Mwab9kyePDBtC1B+Wpb99wDpVL1n7PicgCYjRN33AGHDqXtiONhsG4dzJ3rILCTOQDMxolSKY0CIE0BSfDii7BxY7p5VGB9OQDMxqFyGPT0wH33pX0eFVhfDgCzcWywUcHGjfC978Hq1WkKyWFQLGrm72Vpa2uLrq6uRpdhNq6URwXlIJg7N40I5szxqGA8kPRSRLQNpa1HAGYFUzk99MADcMMNcNddKQQ2bvT0UJF4BGBmJ40KwCOCsapuIwBJHwIeB2YArwI3RcTb/bT7A7Alu/t6RCyopV8zy1ffReNGnT3U05Pq8HpEfdQ0ApD0beBARNwn6W5gckT8t37aHY6IM4Z7fI8AzBqj2tlDoz0quPtu+Na3PPqoxXBGALUGwA7gqojYK2kK8EJEzOqnnQPAbIyq5/RQTw/cdFNah4DUx09+4hAYjnoGwG8j4uxsW8Db5ft92h0FNgFHgfsi4qdDOb4DwKx51CsI+jtLySEwdLkGgKRngfP7eehe4JHKN3xJb0fE5H6OMS0i9ki6CHgOuDoifl2lvyXAEoDp06f/yWuvvTaU52FmdVLPICiPBjwlNHRNNwXU52ceBv4uIp4Y7PgeAZg1r2pB8PGPwxln5PNm7Smh4atnAHwH2F+xCPyhiPhqnzaTgXcj4n1JLcCLwMKI2DbY8R0AZs2vvyCA/P5q73t8jwYGVs8AOAf4CTAdeI10GugBSW3A7RFxm6TLgb8GjpG+f+B7EbFyKMd3AJiNHeU36kOHYOvW/KeHPBoYmroFwGhzAJiNTaM1PeTRwOAcAGbWFAaaHqolDPqOBhwExzkAzKypVJsegvTmvWbNyEKgb7gsXXr86qdF5QAws6ZVfuP+x3+E7u60r5YRQU8PLFyYQmD2bJg3r9gjAQeAmTW9atND554Ln/40fPObQ38T7zslNGVK+o6D9vb86252DgAzGzMqp4eeeQbefDPtH+6ooHycJ56AffvgvPPgc58r3mjAAWBmY1JnJ3z5y/D736e1grLhTO10dqY3/jfeSPdHusYwVg0nAE4Z7WLMzIaqvR02bIC1a9OC7uzZaX93d7oi6fz5aX9Pz8DHKF8+AtL0UlsbLFo08M8VkUcAZta0+lswhrROcOWVaZqn2qig75QQFGM04CkgMxtXqq0TALS0pJFBtUXjzs50llD5Z849Fzo6xu8CsQPAzMat8jrB734Hr7xyfP8558D558PKlSe/uVeeKgrwgQ/AjTcO70yjscJrAGY2bpXXCZ59Fm65Ja0TtLTA/v1pmuj669O+yjn/UilN/SxaBKefDu++C489ltYGOjsb+3waySMAMxvzOjth8eI0zfPWW8f3T54MU6eeOCroOyU0cSI8+SQsGCffVO4RgJkVSnt7Om10w4Y0KmhpSfvffjuNCi6/HGbNSm/+7e2wfv3xM4yOHoUbbkhTQkU7S8gBYGbjRqkEjz56PAg++lGYMAGOHUvrBfPmwdlnwzXXwF/9VXrTh/TF96tWwSc+UawpIU8Bmdm41tGRpocOH4b33ju+X4IPfhA+9rET3/RPOSVdRmKsTgl5CsjMLLNgAfT2wgsvpBHBqaem/RFw8ODJf/EfO5bWCDo66l5q3TkAzKwQ2tthxw7YuTNdKuKss9Jf+9X8+Z/Xr7ZGqSkAJH1BUrekY9nXQFZrd62kHZJ2Zt8dbGbWEKVSOuvnt7+FX/4SLrwwrRP01cSz47mpdQSwFfgc8ItqDSRNAJYB84GLgS9KurjGfs3MatbeDrt2pTOBVqw4vn/SJHj44YaVVTc1BUBEbI+IHYM0uxTYGRG7IuII8GNgYS39mpnlbfHi9Fd/BBw5MnYXgYejHmsA04DKs2t3Z/vMzKyBJg7WQNKzwPn9PHRvRKzJuyBJS4AlANOnT8/78GZmlhk0ACLimhr72ANUXm7pgmxftf6WA8shfQ6gxr7NzKyKekwBrQdmSrpQ0qnAzUABzrA1M2tutZ4G+llJu4HLgJ9JejrbP1XSWoCIOArcCTwNbAd+EhHd1Y5pZmb1MegU0EAiYjWwup/9bwDXVdxfC6ytpS8zM8uXPwlsZlZQDgAzs4JyAJiZFZQDwMysoBwAZmYF5QAwMysoB4CZWUE5AMzMCsoBYGZWUA4AM7OCcgCYmRWUA8DMrKAcAGZmBeUAMDMrKAeAmVlBOQDMzArKAWBmVlC1fiXkFyR1SzomqW2Adq9K2iJpk6SuWvo0M7N81PSVkMBW4HPAXw+h7Scj4q0a+zMzs5zU+p3A2wEk5VONmZnVTb3WAAJ4RtJLkpbUqU8zMxvAoCMASc8C5/fz0L0RsWaI/VwZEXsknQv8XNKvIuIXVfpbAiwBmD59+hAPb2Zmw6WIqP0g0gvAf42IQRd4Jf0P4HBE/K8htO0FXqu5wJFrAZpx3cJ1DU+z1gXNW5vrGp5mquvDEdE6lIa1LgIPStIfA6dExKFs+z8B3xjKzw71SYwWSV0RUfXspkZxXcPTrHVB89bmuoanWesaTK2ngX5W0m7gMuBnkp7O9k+VtDZrdh7wz5I2A/8P+FlE/H0t/ZqZWe1qPQtoNbC6n/1vANdl27uAT9TSj5mZ5c+fBB7Y8kYXUIXrGp5mrQuatzbXNTzNWteAclkENjOzsccjADOzgnIAVGjWaxsNo65rJe2QtFPS3XWo60OSfi7pX7P/Tq7S7g/Za7VJUsco1jPg85d0mqTHs8c7Jc0YrVqGWdeXJPVWvEa31amuhyTtk7S1yuOS9P2s7pclXdIkdV0l6WDF6/X1OtVVkvS8pG3Z/49f6adNQ16zEYsI37Ib8B+AWcALQNsA7V4FWpqpLmAC8GvgIuBUYDNw8SjX9W3g7mz7buBbVdodrsNrNOjzB5YCP8i2bwYeb5K6vgTcX6/fp4p+5wGXAFurPH4d8BQgYC7Q2SR1XQX8XQNerynAJdn2mcAr/fxbNuQ1G+nNI4AKEbE9InY0uo6+hljXpcDOiNgVEUeAHwMLR7m0hcAj2fYjwA2j3N9AhvL8K+t9Arhao38hq0b8uwxJpE/jHxigyULg0UjWAWdLmtIEdTVEROyNiA3Z9iFgOzCtT7OGvGYj5QAYmWa8ttE0oKfi/m5O/uXM23kRsTfb/g3pMx/9+SNJXZLWSRqtkBjK8//3NhFxFDgInDNK9QynLoAbsymDJySVRrmmoWrE79RQXSZps6SnJM2ud+fZ9OEcoLPPQ838mp1k1D8J3GzqfW2jOteVu4HqqrwTESGp2illH85er4uA5yRtiYhf513rGPa3wI8i4n1J/5k0SvlUg2tqZhtIv1OHJV0H/BSYWa/OJZ0BPAncFRHv1Kvf0VC4AIiIa3I4xp7sv/skrSYN82sKgBzq2gNU/uV4QbavJgPVJelNSVMiYm82zN1X5Rjl12tXdt2oOaR58TwN5fmX2+yWNBE4C9ifcx3DrisiKmtYQVpbaQaj8jtVq8o33YhYK+kBSS1Rh+8bkTSJ9Ob/w4hY1U+TpnzNqvEU0DBJ+mNJZ5a3Sdc26vdshTpbD8yUdKGkU0mLnKN2xk2mA7g1274VOGmkImmypNOy7RbgCmDbKNQylOdfWe/ngeciW7kbRYPW1WeOeAFpbrkZdACLsjNb5gIHK6b8GkbS+eW1G0mXkt7HRjvIyfpcCWyPiO9WadaUr1lVjV6FbqYb8FnSnN37wJvA09n+qcDabPsi0pkcm4Fu0hRNw+vK7l9HOjPh13Wq6xzgH4B/BZ4FPpTtbwNWZNuXA1uy12sLsHgU6znp+ZMuPLgg2/4j4P8CO0nXpbqoTr9Xg9X1P7Pfpc3A88DH6lTXj4C9wO+z36/FwO3A7dnjApZldW9hgDPj6lzXnRWv1zrg8jrVdSVp/e9lYFN2u64ZXrOR3vxJYDOzgvIUkJlZQTkAzMwKygFgZlZQDgAzs4JyAJiZFZQDwMysoBwAZmYF5QAwMyuo/w9oj1uhpQW7WwAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f1bd0250050>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0537431649864\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f1b7a579910>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0363325290382\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f1bd01fff90>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0272595621645\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f1b79d1f1d0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0217338036746\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f1b7a56ee50>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loss: 0.0180306117982\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f1c0c0193d0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "graph = tf.Graph()\n",
    "sess = tf.Session(graph=graph)\n",
    "X, L_1, X_2, L_2, Y = train(graph, sess)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "with graph.as_default():\n",
    "    with sess.as_default():\n",
    "        l_1_a, x_2_a, l_2_a, y_a = sess.run([L_1, X_2, L_2, Y], feed_dict={X: x_a})\n",
    "        l_1_b, x_2_b, l_2_b, y_b = sess.run([L_1, X_2, L_2, Y], feed_dict={X: x_b})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f1bd0184050>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f1c747b6b50>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f1bd8044c10>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f1b7af60ed0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot1d(x_a, x_b)\n",
    "plot2d(l_1_a, l_1_b)\n",
    "plot2d(x_2_a, x_2_b)\n",
    "plot2d(l_2_a, l_2_b)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "plt.xlim(x_min,x_max)\n",
    "plt.ylim(y_min,y_max)\n",
    "plt.scatter(x_a, [0]*len(x_a))\n",
    "plt.scatter(x_b, [0]*len(x_b), c='r')\n",
    "plt.show()\n",
    "\n",
    "plt.xlim(x_min*5,x_max*5)\n",
    "plt.ylim(y_min*5,y_max*5)\n",
    "plt.scatter(y_a.T[0], y_a.T[1])\n",
    "plt.scatter(y_b.T[0], y_b.T[1], c='r')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "feed_dict = {X: x_a}\n",
    "\n",
    "l_1_a = sess.run(L_1, feed_dict=feed_dict)\n",
    "x_2_a = sess.run(X_2, feed_dict=feed_dict)\n",
    "y_a = sess.run(Y, feed_dict=feed_dict)[0]\n",
    "\n",
    "feed_dict[X] = x_b\n",
    "\n",
    "l_1_b = sess.run(L_1, feed_dict=feed_dict)\n",
    "x_2_b = sess.run(X_2, feed_dict=feed_dict)\n",
    "y_b = sess.run(Y, feed_dict=feed_dict)[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "plt.xlim(x_min*3,x_max*3)\n",
    "plt.ylim(y_min*3,y_max*3)\n",
    "plt.scatter(l_1_a[0], l_1_a[1])\n",
    "plt.scatter(l_1_b[0], l_1_b[1], c='r')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "plt.xlim(x_min,x_max)\n",
    "plt.ylim(y_min,y_max)\n",
    "plt.scatter(x_2_a[0], x_2_a[1])\n",
    "plt.scatter(x_2_b[0], x_2_b[1], c='r')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "plt.xlim(x_min,x_max)\n",
    "plt.ylim(y_min,y_max)\n",
    "plt.scatter(y_a, [0]*len(y_a))\n",
    "plt.scatter(y_b, [0]*len(y_b), c='r')\n",
    "plt.show()"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 2",
   "language": "python",
   "name": "python2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}