Wavelet_Network Test2.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/edoardobucheli/anaconda3/lib/python3.6/site-packages/h5py/__init__.py:36: FutureWarning: Conversion of the second argument of issubdtype from `float` to `np.floating` is deprecated. In future, it will be treated as `np.float64 == np.dtype(float).type`.\n",
      "  from ._conv import register_converters as _register_converters\n",
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "import tensorflow as tf\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "from keras.datasets import mnist\n",
    "\n",
    "from Utilities import make_oh"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "(x_train, y_train), (x_test, y_test) = mnist.load_data()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_train_2 = make_oh(y_train)\n",
    "y_test_2 = make_oh(y_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "x = tf.placeholder(dtype = tf.float32, shape = (None,28,28), name = 'x_input')\n",
    "y = tf.placeholder(dtype = tf.float32, shape = (None,10), name = 'labels')\n",
    "\n",
    "d_1 = tf.get_variable(name = 'dilation_1',dtype = tf.float32, initializer=tf.constant(0.2, dtype = tf.float32))\n",
    "t_1 = tf.get_variable(name = 'translation_1', dtype = tf.float32, initializer=tf.constant(0.0, dtype = tf.float32))\n",
    "\n",
    "#d_2 = tf.get_variable(name = 'dilation_2',dtype = tf.float32, initializer=tf.constant(0.2, dtype = tf.float32))\n",
    "#t_2 = tf.get_variable(name = 'translation_2', dtype = tf.float32, initializer=tf.constant(0.0, dtype = tf.float32))\n",
    "\n",
    "d_3 = tf.get_variable(name = 'dilation_3',dtype = tf.float32, initializer=tf.constant(0.2, dtype = tf.float32))\n",
    "t_3 = tf.get_variable(name = 'translation_3', dtype = tf.float32, initializer=tf.constant(0.0, dtype = tf.float32))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "def wavelet_act(x,d,t):\n",
    "    x = tf.multiply(d,(x-t))\n",
    "    return x*tf.exp(-0.5*x**2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "def mexican_hat_wt(x,d,t):\n",
    "    x = (x-t)\n",
    "    tf_pi = tf.constant(np.pi,dtype=tf.float32)\n",
    "    #f = tf.constant(0.1, dtype = tf.float32)\n",
    "    #return (1-x**2)*tf.exp(-x**2)\n",
    "    return (1-2*(tf_pi**2)*(d**2)*(x**2))*tf.exp(-(tf_pi**2)*(d**2)*x**2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "#check_wt = mexican_hat_wt(x,d_1,t_1)\n",
    "\n",
    "x_new = tf.expand_dims(x,axis = -1)\n",
    "\n",
    "z_1 = tf.layers.conv2d(x_new,32,[3,3])\n",
    "a_1 = mexican_hat_wt(z_1,d_1,t_1)\n",
    "\n",
    "#z_2 = tf.layers.conv2d(a_1,64,[3,3])\n",
    "#a_2 = mexican_hat_wt(z_2,d_2,t_2)\n",
    "\n",
    "max_pool_1 = tf.layers.max_pooling2d(a_1,[2,2],[2,2])\n",
    "\n",
    "flat = tf.layers.flatten(max_pool_1)\n",
    "\n",
    "fc_z_1 = tf.layers.Dense(32,activation=None)(flat)\n",
    "fc_a_1 = mexican_hat_wt(fc_z_1,d_3,t_3)\n",
    "\n",
    "\n",
    "logits = tf.layers.Dense(10)(fc_a_1)\n",
    "y_hat = tf.argmax(tf.nn.softmax(logits),axis = 1)\n",
    "\n",
    "acc = tf.reduce_mean(tf.cast(tf.equal(y_hat,tf.argmax(y,axis = 1)),dtype = tf.float32))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "loss = tf.nn.softmax_cross_entropy_with_logits_v2(labels = y, logits=logits)\n",
    "optimizer = tf.train.AdamOptimizer()\n",
    "\n",
    "tr_step = optimizer.minimize(loss)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "iters = 500"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "with tf.Session() as sess:\n",
    "\n",
    "    sess.run(tf.global_variables_initializer())\n",
    "\n",
    "    for i in range(iters):\n",
    "        this_acc,_ = sess.run([acc,tr_step],feed_dict = {x:x_train,y:y_train_2})\n",
    "        \n",
    "        if i%5 == 0:\n",
    "            this_test_acc = sess.run(acc, feed_dict = {x:x_test,y:y_test_2})\n",
    "            this_d, this_t = sess.run([d_1,t_1])\n",
    "            print('It {:d}:\\tTrain Acc {:02.2f}%\\n\\tTest Acc {:02.2f}%\\n'.format(i,this_acc*100,this_test_acc*100))  \n",
    "            print('d: {:01.2f}, t: {:01.2f}\\n'.format(this_d,this_t))\n",
    "        elif i == iters-1:\n",
    "            this_test_acc = sess.run(acc, feed_dict = {x:x_test,y:y_test_2})\n",
    "            this_d, this_t = sess.run([d_1,t_1])\n",
    "            print('Final It:\\tTrain Acc {:02.2f}%\\n\\tTest Acc {:02.2f}%'.format(this_acc*100,this_test_acc*100))  \n",
    "            print('d: {:01.2f}, t: {:01.2f}\\n'.format(this_d,this_t))\n",
    "    \n",
    "    #check_values = np.linspace(-10,10,num = 784)\n",
    "    #check_values = np.reshape(check_values,(1,784))\n",
    "    \n",
    "    #edo = sess.run(check_wt,feed_dict = {x:check_values})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x105f71ba8>]"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAX8AAAD8CAYAAACfF6SlAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMi4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvhp/UCwAAIABJREFUeJzt3Xl0m9d95vHvjwu4kyJFal8o27Id2Wm90LITp4lT24nt9lhJJos9Wdw0qSdN3JlO2s44Jz2enuR0pst02snUbaosk7Vx3E5aazJKHMdJmziJZcuJF9myLFqSLWqluIsrSP7mD7ygYQggQeAFCBDP5xweAS8u3nsJQg8u7nvf+5q7IyIi5aViqRsgIiKFp/AXESlDCn8RkTKk8BcRKUMKfxGRMqTwFxEpQwp/EZEypPAXESlDCn8RkTJUtdQNSKe9vd07OzuXuhkiIiXliSeeOOPuHQuVK9rw7+zsZO/evUvdDBGRkmJmL2VSTsM+IiJlSOEvIlKGFP4iImVI4S8iUoYU/iIiZSiU8DezL5rZaTPbl+ZxM7PPmFm3mT1tZleEUa+IiGQnrJ7/l4Cb5nn8ZmBr8HMn8Lch1SsiIlkIJfzd/UdA/zxFdgBf8ZhHgRVmtjaMukUKbc+hPr799HF0CVQpZYU6yWs9cDThfk+w7URiITO7k9g3AzZt2lSgpolk7tnjQ7xn56MATL5rln9z5YYlbpFIdgp1wNdSbDun2+TuO929y927OjoWPDtZpOC+8djL1FRVsK6lli/99MhSN0cka4UK/x5gY8L9DcDxAtUtEpofvXCGN1zQznuv2cwzx4boH51a6iaJZKVQ4b8L+EAw6+caYMjdTyz0JJFicuTMKC/3j/Gmizq4qrMNgCdeGljiVolkJ5QxfzP7BnAd0G5mPcB/AaoB3P2zwG7gFqAbGAM+GEa9IoX05NFBALZvaaNzZQPVlcYTLw1w47bVS9wykcULJfzd/fYFHnfgY2HUJbJUDpwaobrSOK+9kUhVBdvWNvPkUfX8pTTpDF+RDB04OTIX/ABbVzfxYu/oErdKJDsKf5EMHTg5wkVrmubun9fRQO/IJCMT0SVslUh2FP4iGRibmubY4DgXrm6c23Z+R+z2IfX+pQQp/EUycLR/HIBNKxvmtp3fEbt96MzZJWmTSC4U/iIZONo/BsDG1rq5bZvaGqgwOKyev5Qghb9IBnoGYuG/obV+blukqoI1zbUcG5xYqmaJZE3hL5KBowPj1FVX0t4YedX2dSvqOD44vkStEsmewl8kA0f7x9jQWofZq5epWreijuNDCn8pPQp/kQz0DIyzPmG8P27tilpODE4wO6vlnaW0KPxFMnB6ZIK1LbXnbF+/oo6pmVnOjE4uQatEsqfwF1nA9MwsfaNTdDSdG/7rWmLfBo7roK+UGIW/yALOnJ3CHVY11Zzz2Jrg28BJjftLiVH4iyzg9EisV58q/OPbes9qXX8pLQp/kQWcHo6N569qPnfYp60hghn0jmjMX0qLwl9kAafm6flXVVawsiGi8JeSo/AXWUC859+RIvwB2htrOHNW4S+lReEvsoDTI5OsbIhQXZn6v0tHU416/lJyFP4iC+gdmUjb6wfoaFT4S+kJJfzN7CYzO2Bm3WZ2d4rHN5nZD83sF2b2tJndEka9IoVwemQy5cHeuPamGnrPThK7WqlIacg5/M2sErgXuBnYBtxuZtuSiv0hcL+7Xw7cBvxNrvWKFMrp4cmUB3vjOhprmJqeZWRyuoCtEslNGD3/7UC3ux9y9yngPmBHUhkHmoPbLcDxEOoVybvZWefM2cn5h33ic/019CMlJIzwXw8cTbjfE2xL9EfA+8ysB9gN/E4I9Yrk3cjENNOzzsqGSNoy7Y0Kfyk9YYS/pdiWPPh5O/Ald98A3AJ81czOqdvM7jSzvWa2t7e3N4SmieRmYCx25m7bPOEf7/lruqeUkjDCvwfYmHB/A+cO63wIuB/A3X8G1ALtyTty953u3uXuXR0dHSE0TSQ3/UH4t2YQ/vHzAURKQRjh/ziw1cy2mFmE2AHdXUllXgauBzCz1xALf3XtpegNjAY9//r04d9SV02FvfItQaQU5Bz+7j4N3AU8COwnNqvnWTP7lJndGhT7PeC3zOwp4BvAb7jmxUkJ6B9deNinssJYUR9R+EtJqQpjJ+6+m9iB3MRt9yTcfg64Noy6RAopHugr6qvnLbeivpqB0WghmiQSCp3hKzKP/tEo1ZVGY838/aS2+sjctwSRUqDwF5nHwOgUrfWRcy7cnqy1QcM+UloU/iLz6B+bmne8P661vlrhLyVF4S8yj8GxWM9/Ia0NEQZGo1rfR0qGwl9kHv2jmfX82+ojTM3MMjY1U4BWieRO4S8yj4GxKK0N88/0Aea+Heigr5QKhb9IGjOzzuDY1LwneMXFzwDWuL+UCoW/SBrD41Fmff6lHeLagm8HA2Oa6y+lQeEvksbcuj4Z9PxXBGUGNOwjJULhL5JGPMgz6vnXa9hHSovCXySN/gwWdYtrji/upp6/lAiFv0gag8H4fSazfSorjJa66rmhIpFip/AXSaM/gwu5JGqtj8x9YIgUO4W/SBoDo1NEqiqoq67MqHxzXTVD4wp/KQ0Kf5E0hsajrKirXnBRt7iWumqGFf5SIhT+ImkMjUdprlt4vD+uRT1/KSEKf5E0hsajtCj8ZZlS+IukMTyx+PAfnpjWyp5SEkIJfzO7ycwOmFm3md2dpsy7zew5M3vWzP4+jHpF8imbnv/MrHN2cjqPrRIJR87X8DWzSuBe4EagB3jczHYF1+2Nl9kKfAK41t0HzGxVrvWK5NvQWJTm2sz/i8Q/KIbGozTVZv6hIbIUwuj5bwe63f2Qu08B9wE7ksr8FnCvuw8AuPvpEOoVyZvZWWdkcnpRPf/mhPAXKXZhhP964GjC/Z5gW6ILgQvN7Cdm9qiZ3RRCvSJ5MzI5jTuLnu0DCn8pDTkP+wCpJkEnH/GqArYC1wEbgB+b2aXuPviqHZndCdwJsGnTphCaJpKd+Hz9xY75Jz5XpJiF0fPvATYm3N8AHE9R5gF3j7r7YeAAsQ+DV3H3ne7e5e5dHR0dITRNJDtD2YR/vXr+UjrCCP/Hga1mtsXMIsBtwK6kMv8MvBnAzNqJDQMdCqFukbyIB7iGfWS5yjn83X0auAt4ENgP3O/uz5rZp8zs1qDYg0CfmT0H/BD4A3fvy7VukXzJpuffEKmkssIU/lISwhjzx913A7uTtt2TcNuBjwc/IkUvmzF/M9NZvlIydIavSArZ9Pzj5YfGdZKXFD+Fv0gKQ+NRKiuM+khmyznHaVlnKRUKf5EU4ks7ZLqcc5yGfaRUKPxFUhieWNzZvXFa019KhcJfJIXFruUf11JXpZ6/lASFv0gKQ+OLW9QtLj7so2Wdpdgp/EVSGF7kcs5x8WWdR6dm8tAqkfAo/EVSyCX8QWf5SvFT+IskcfdFX8glbi78xxT+UtwU/iJJxqZmmJ71rMJfa/pLqVD4iyTJZlG3OA37SKlQ+IskGZ7IbmmHxOdorr8UO4W/SJL4eL0O+MpypvAXSZLtom4AjTVVWtZZSoLCXyTJ3Jh/7eLD38xortVZvlL8FP4iSYYnYksyZ9Pzh9iB4vhxA5FipfAXSTI0HsUMmrJY3gG0sqeUBoW/SJLh8ShNNVVUVCxuOee45lqt7CnFT+EvkiTbFT3j1POXUhBK+JvZTWZ2wMy6zezuecq908zczLrCqFckH7Jd2iGuua5q7riBSLHKOfzNrBK4F7gZ2AbcbmbbUpRrAv49sCfXOkXyKdtF3eJ0KUcpBWH0/LcD3e5+yN2ngPuAHSnKfRr4M2AihDpF8ibnnn9tNVPTs0xEtayzFK8wwn89cDThfk+wbY6ZXQ5sdPdvz7cjM7vTzPaa2d7e3t4QmiayeMMT0azm+MfFjxdouqcUszDCP9WUiLnLGJlZBfCXwO8ttCN33+nuXe7e1dHREULTRBZvaDxKS31uB3xB6/tIcQsj/HuAjQn3NwDHE+43AZcC/2JmR4BrgF066CvFaHJ6honobFaXcIyLP3doXAd9pXiFEf6PA1vNbIuZRYDbgF3xB919yN3b3b3T3TuBR4Fb3X1vCHWLhGokmKWT61RPUM9filvO4e/u08BdwIPAfuB+d3/WzD5lZrfmun+RQhrOYVG3OI35SynI/rttAnffDexO2nZPmrLXhVGnSD7ksqhbnJZ1llKgM3xFEgzPDftk3y+KrwmkYR8pZgp/kQTDIfT8a6oqqa2u0Fm+UtQU/iIJcrmQS6KWuuq5K4KJFCOFv0iC+EHaXGb7QLCypw74ShFT+IskGB6fJlJZQU1Vbv81tLKnFDuFv0iC4YkozXVVmGW3ln+cruYlxU7hL5JgaDy3dX3i1POXYqfwF0kwnOOFXOKaa6sY1vIOUsQU/iIJhiemwwn/YNhndtYXLiyyBBT+IglGxqM5LeoW11JXjTucnVLvX4qTwl8kQa7X742LHzfQWb5SrBT+IgF3Z3git6t4xTVrfR8pcgp/kcBEdJbojIcy2ye+NpAO+kqxUviLBF45uzecMX9Qz1+Kl8JfJBDGcs5xc2P+OtFLipTCXyQQxoVc4pp1NS8pcgp/kUBYi7oBNNVUYabwl+Kl8BcJvDLsk/uYf0WF0VRTpTX9pWiFEv5mdpOZHTCzbjO7O8XjHzez58zsaTN72Mw2h1GvSJjiM3PC6PkDtNRrfR8pXjmHv5lVAvcCNwPbgNvNbFtSsV8AXe7+S8A/An+Wa70iYQvjKl6JmmurNewjRSuMnv92oNvdD7n7FHAfsCOxgLv/0N3HgruPAhtCqFckVMMTUeqqK4nkuJZ/nFb2lGIWxrt8PXA04X5PsC2dDwHfCaFekVDFlnbIfbw/TlfzkmIWxjs91VUvUi5laGbvA7qAN6V5/E7gToBNmzaF0DSRzA2PT4c25AOxk8XU85diFUbPvwfYmHB/A3A8uZCZ3QB8ErjV3SdT7cjdd7p7l7t3dXR0hNA0kcyFta5PXEtdtZZ3kKIVRvg/Dmw1sy1mFgFuA3YlFjCzy4G/Ixb8p0OoUyR0sUs4htjzr61mPDrD1PRsaPsUCUvO4e/u08BdwIPAfuB+d3/WzD5lZrcGxf4caAT+wcyeNLNdaXYnsmSGQlrLP66lXks8SPEK5Z3u7ruB3Unb7km4fUMY9Yjk0/B4OFfxiktc07+9sSa0/YqEQWf4igCzs85IHsb8QSt7SnFS+IsAo1PTzHp4J3hBwpr+WuJBipDCX4SEdX1CnOevnr8UM4W/CAnr+oTZ89d1fKWIKfxFeGVGTphj/rqOrxQzhb8IicM+4YV/bbBOkKZ6SjFS+IsQ/oqecbGVPXXAV4qPwl+EV2bkhHnAF6Clrkpj/lKUFP4ivNLzbwq751+nlT2lOCn8RYiN+TfVVFFZkWqR2uxpTX8pVgp/EWI9/zAP9sbpal5SrBT+IsDgeJQV9XkIf63pL0VK4S8CDI5N5SX8W+qqGZ6Yxj3l9Y1ElozCX4Sg518XCX2/zbXVzMw6Y1Mzoe9bJBcKfxFgaCw/wz5a30eKlcJfyp6753HMXxd0keKk8Jeyd3ZymplZz8uwz1zPf0zhL8VF4S9lbzAI5pZ89PzjK3tqTX8pMqGEv5ndZGYHzKzbzO5O8XiNmX0zeHyPmXWGUa9IGOLhvyIf8/yD5SI05i/FJufwN7NK4F7gZmAbcLuZbUsq9iFgwN0vAP4S+NNc6xUJy+D4FAAr6vM47KPwlyITRs9/O9Dt7ofcfQq4D9iRVGYH8OXg9j8C15tZuOfRi2Qp3vNvzcOwT1NtNWYwNDYV+r5FchHGEobrgaMJ93uAq9OVcfdpMxsCVgJnQqj/VQbHpnjP3z36qm0Lfcwkfg4lF01+buJ9Syp9TtlzK8q8njTty+y5r2pkxs9d8PeZp575XuNIZQXNddU011bRUh9hw4o6tnQ0cOHqplAvnpKtwfH8jflXVhgtddUMFMkB34noDC+cGuFQ7ygv948xMDbF0FiU4Ykok9OzTM8407OzRGec6MwssynOTcv0hDWd15a9i9c28T9vuzyvdYQR/qn+2yf/2TMpg5ndCdwJsGnTpqwaU1FhbGlvSKjk1dUkvyF9nseSm5j4eHLR5P8Q5z6e+XPTPS/23AV+H19E2cRt/uqWLe51m/93n5qe5YXTIwyNRRmZnJ7bV4XBa9e3cMNrVvPuqzayurmWpRDvlefrg6i1PsLAEvb8Ryai/POTx/nuvhM8fmSAqenZuceaaqpoqa+mubaaSFUF1ZVGVUUFtdVGdWVF2oXuUm1N1QFI7lBIZta21OW9jjDCvwfYmHB/A3A8TZkeM6sCWoD+5B25+05gJ0BXV1dW/Ybm2mo++/4rs3mqFMD0zCzHBsc5dGaUX7w8yE+6z/AXD73AXz18kPdfs5nfvWFrXsbe5zM4FqU+UklNVWVe9t9aX70k4T81Pcv//slh/tcPujk7Oc2Fqxv5wDWb6eps5fyORjatrM/b7yzFL4zwfxzYamZbgGPAbcC/TSqzC7gD+BnwTuAHrsVOylJVZQWbVzaweWUDb75oFR+/8UKOnBll548P8ZWfHWH3Mye4971XcFVnW8HaFFvaIX/DT631EU4MTeRt/6kc7R/j3331CZ47Mcz1F6/id67fyi9vaDlnCFHKV84HfN19GrgLeBDYD9zv7s+a2afM7Nag2BeAlWbWDXwcOGc6qJSvzvYG/uvbX8uuu95AfaSS935uD99/7lTB6h8ci9KSx28bK+ojDBaw53/g5Ahvu/cn9AyMsfP9V/KF37iKyzauUPDLq4RyzTp33w3sTtp2T8LtCeBdYdQly9el61t44GNv4ANf3MNvf/0Jvvahq7n6vJV5r3dofCovM33i2hoKd8C3Z2CM937+Uaoqjfs/ci3ndzQWpF4pPTrDV4pKS301X/nQ1Wxsq+ejX/85J4bG817nYJ4WdYtbUR9hPDrDRDS/K3tORGf4yNeeYDI6y9c/fI2CX+al8Jei01JXzec+0MV4dIZPfOuZvK+FPzgepSUP6/rEtQZDSvk+6PuZhw+y79gwf/mey7hglYJf5qfwl6J0fkcj/+mtF/EvB3p54MnkyWPhcfe8Lecc19YQ2/fAaP6Gfp47Pszf/egQ77pyAzdsW523emT5UPhL0frA6zp57foW/vzBA3kbMhmbmmFqZjavs31WFKDn/9++s5/m2io++WuvyVsdsrwo/KVoVVQYn7j5Yo4NjvPVn72UlzriZ/fms+ef72Gfn754hh8fPMPH3nxBwc+RkNKl8Jei9voL2nnDBe187seHmJwOv/c/MJq/Rd3i4jOJ8jXj569/0M3q5hred83mvOxflieFvxS9O994HqdHJtmVh7H/viD8Vzbkd54/vPJBE6Znjw/x0xf7+OC1W6it1tm6kjmFvxS9X9nazsVrmvjCI4dDn/nTPzoJQFsewz9SVUFjTVVehn2++MgR6iOV3H5VdmthSflS+EvRMzPee81mnj85wr5jw6Huu+9svOdfE+p+k62or55bOjosZyen2f3MCXZctj4vK5LK8qbwl5Jw6y+vo6aqgm/ufTnU/Q6MTVFVYXNX3MqXtoYI/SEP++x+5gTj0RneeeX6UPcr5UHhLyWhpa6amy9dwwNPHg/1wG//6BStDZG8r3uTj/V9vvXzHra0N3DFptZQ9yvlQeEvJWPHZesZmZjmp919oe2z7+wUbQWYHhlb1jm8YZ+egTEePdTPOy5frwXbJCsKfykZr79gJU01Vex+5kRo++wfncrrwd64sC/o8t19J4HYB6JINhT+UjJqqiq5YdtqHtp/iujM7MJPyED/6BRtjYUJ/5GJ6VddRSsX33v2FBevaWLTyvpQ9iflR+EvJeWmS9cwOBbl0UPhDP30jU7ldY5/3MrgAyaMg779o1Psfamft2gNH8mBwl9Kypsu7KCuupLvPZv7xV6iM7MMjUcLMuzT3hibSnrm7GTO+3p4/ylmHW7ctibnfUn5UvhLSamtruT156/kRwd7c95XfAy+EOHf0RSrozeE8H/ouVOsbanl0vXNOe9LypfCX0rOGy/s4KW+MY6cGc1pP/EhmIL2/EdyC/+ZWeenL/Zx3UWrNMtHcqLwl5Lzpgs7APjXF3Lr/fefLXz49+U45r//xDBnJ6e55rzCXeBelqecwt/M2szsITM7GPx7ztkmZnaZmf3MzJ41s6fN7D251CnS2d7AprZ6fpRj+J8ZLczSDgANNVXUVVfm3PPfc7gfgO1bFP6Sm1x7/ncDD7v7VuDh4H6yMeAD7n4JcBPwV2a2Isd6pcy96cIOfvpiX05n+/YGQbyqKf/hD9DeFMn5gO+eQ31saqtnbUtdSK2ScpVr+O8Avhzc/jLwtuQC7v6Cux8Mbh8HTgMdOdYrZe4NW9sZj87wdM9Q1vs4PTJBpLIirxdySdTeWMOZs9kP+8zOOo8f6edq9folBLmG/2p3PwEQ/LtqvsJmth2IAC+mefxOM9trZnt7e3OfzSHL1/bOWADuyWG+f+/wJB1NNQU7cBoL/+x7/t29ZxkYi2rIR0KxYPib2ffNbF+Knx2LqcjM1gJfBT7o7ilPc3T3ne7e5e5dHR36ciDptTZEuHhN09wYeDZ6z07SXqAhH4D2xtyGfeK/69VbVobVJCljC65j6+43pHvMzE6Z2Vp3PxGE++k05ZqB/wf8obs/mnVrRRJcvaWN+/f2EJ2Zpbpy8V9iTw9PFnR5hPbGGvpHp5iZdSorFv9t47HD/axprmVjm8b7JXe5DvvsAu4Ibt8BPJBcwMwiwD8BX3H3f8ixPpE5V5+3Mqdx/9MjEwU72Aux8J/17C7k7u7sOdTH1ee1aX6/hCLX8P8T4EYzOwjcGNzHzLrM7PNBmXcDbwR+w8yeDH4uy7Fekbmx7z2HFz/uPzU9y8BYlFVNtWE3K62O4IOmN4vpni/1jXF6ZFLj/RKanC5f5O59wPUptu8FPhzc/hrwtVzqEUmlvbGGrasa2XOon49et7jnxsfeOwrY81/dHPugOTk8wWvWLm5phsfmxvsV/hIOneErJe3q89rYe6Sf6UUu8Xy6wHP8Ada2BOE/NLHo5+453M/KhgjndzSG3SwpUwp/KWnbt6xkdGqG/SdGFvW808OxAF7VXLjw72iqocLgRFbh38f2LRrvl/Ao/KWkxef7P3ZkcVM+46trFnLMv7qygo6mGk4OjS/qeccGx+kZGNd4v4RK4S8lbU1LbOrj44uc739qaIIKe+UiK4WypqVu0T3/x7Wej+SBwl9K3lWb23j8SD/unvFzegbHWd1cm9X5AblY01yz6DH/PYf7aaqt4uI1Wr9fwqPwl5J31ZY2+kanOLyI9f2PDYyzfkXhT5Za21KXRfj3cVVnW1Ynhomko/CXkndVMO7/+CLG/Y8PjbNuCcJ/TUstI5PTnJ2czqh878gkh3pHNcVTQqfwl5J3fkcDKxsiPHZ4IKPyM7POicEJ1rcuRc9/cdM94x9oGu+XsCn8peSZGV2drRn3/E+PTDA960sy7LOmeXHh/9jhfuqqK7l0fUs+myVlSOEvy8JVnW283D/GqeGFQ/XYQGyq5VL0/ONDTT0DYxmV33O4nys3txb8wLQsf3pHybIQH/d/LIMpn8cGY+G/YQl6/utW1FFdaRzpWzj8h8aiPH9yWOP9khcKf1kWLlnXTH2kkr0ZDP30LGHPv7LC2NhWz0t9C89Mik1f1Xi/5IfCX5aFqsoKrtjUymNHFj7oe2xwnNb6auojOa1rmLUtKxsympb62JF+IpUV/PJGXfJawqfwl2Xjqs42nj85zNB4dN5yL/WNsqmtcBdxSbZ5ZQMv9Y0teFLansP9XLZxBbXVlQVqmZQThb8sG1dtacUdfv7S/L3/Q72jS7o6Zmd7PePRmbmVRVM5OznNvmNDGvKRvFH4y7Jx+cZWqips3kXeRienOTE0wXkdDQVs2at1rozVfWSeoZ/HDvcxM+u87nxdr1fyQ+Evy0ZdJDYffr5F3uJj7ectZc8/CP+X5pnx89PuPiJVFVy5ubVQzZIyo/CXZWX7ljae7hliIjqT8vEXe88CLOmwz7oVtUQqK+bakspPXuyja3Orxvslb3IKfzNrM7OHzOxg8G/aboqZNZvZMTP761zqFJnPVZ1tTM3M8tTRwZSPP39yhKoKo7N96Q74VlVWsHV1I8+dGE75eN/ZSfafGObaC9oL3DIpJ7n2/O8GHnb3rcDDwf10Pg38a471icxr+5bY6pf/+kJvysf3HRviwtVN1FQtbY/6knXNPHd8OOWMn0cPxYatXq/xfsmjXMN/B/Dl4PaXgbelKmRmVwKrge/lWJ/IvFrqquna3MoPnj99zmPuzr5jQ1y6funXxd+2tpm+0SlODZ874+eR7l6aaqp4rdbzkTzKNfxXu/sJgODfVckFzKwC+AvgD3KsSyQjv3rxKp4/OcLxwVdfLvH40AQDY9GiCNVLgjY83fPq4amZWeeh507xpos6qNJ6PpJHC767zOz7ZrYvxc+ODOv4KLDb3Y9mUNedZrbXzPb29qb+2i6ykF+9ONYHSe797znUB8BlG5d+Bs1r17cQqaqYG+KJe/xIP2fOTnHzpWuXqGVSLhY8v93db0j3mJmdMrO17n7CzNYC537XhtcBv2JmHwUagYiZnXX3c44PuPtOYCdAV1dX5tfkE0lwwapGzutoYNdTx3nfNZvntj/SfYbW+mouWbf0wz611ZV0bW7lZ8EHUtx3952kpqqC6y7qWKKWSbnI9XvlLuCO4PYdwAPJBdz9ve6+yd07gd8HvpIq+EXCYma84/L1PHa4n6P9sbn0s7POIwfP8PoL2qkoksshvv78lew/MTy3tv9EdIb/+9Rxrruog4aapVl3SMpHruH/J8CNZnYQuDG4j5l1mdnnc22cSLbeccUGqiqMLzxyGIAfHezl9Mgkb71kzRK37BW//kvrAPjWL3oA+PbTJ+gbneL913QuYaukXOQU/u7e5+7Xu/vW4N/+YPted/9wivJfcve7cqlTJBPrVtTx9svX843HXub5k8N84ZHDtDdGuKmIwr+zvYHtW9r4+qMvc3p4gv/+4AG2rW3m2gs0xVPyT9MJZNltav2eAAAGt0lEQVT6g5suormuml/7zCP8+OAZPvbmC4hUFddb/j/ecCHHh8Z53Z/8gN6zk/zx2y/FrDiGpWR508CiLFurmmr5+w9fzed/fJhL1jfz/oSDv8XideevZOf7u/juvpO8/fL1XL5p6WciSXmwhdYUXypdXV2+d+/epW6GiEhJMbMn3L1roXLF9R1YREQKQuEvIlKGFP4iImVI4S8iUoYU/iIiZUjhLyJShhT+IiJlSOEvIlKGivYkLzPrBV7KYRftwJmQmhMmtWtx1K7FUbsWZzm2a7O7L7gmeNGGf67MbG8mZ7kVmtq1OGrX4qhdi1PO7dKwj4hIGVL4i4iUoeUc/juXugFpqF2Lo3Ytjtq1OGXbrmU75i8iIukt556/iIikUdLhb2bvMrNnzWzWzLqSHvuEmXWb2QEze2ua528xsz1mdtDMvmlmkTy08Ztm9mTwc8TMnkxT7oiZPROUy/uFDMzsj8zsWELbbklT7qbgNew2s7sL0K4/N7PnzexpM/snM1uRplxBXq+Ffn8zqwn+xt3Be6kzX21JqHOjmf3QzPYH7///kKLMdWY2lPD3vSff7QrqnffvYjGfCV6vp83sigK06aKE1+FJMxs2s99NKlOQ18vMvmhmp81sX8K2NjN7KMihh8ws5RV9zOyOoMxBM7sj58a4e8n+AK8BLgL+BehK2L4NeAqoAbYALwKVKZ5/P3BbcPuzwG/nub1/AdyT5rEjQHsBX7s/An5/gTKVwWt3HhAJXtNteW7XW4Cq4PafAn+6VK9XJr8/8FHgs8Ht24BvFuBvtxa4IrjdBLyQol3XAd8u1Psp078LcAvwHcCAa4A9BW5fJXCS2Fz4gr9ewBuBK4B9Cdv+DLg7uH13qvc80AYcCv5tDW635tKWku75u/t+dz+Q4qEdwH3uPunuh4FuYHtiAYtdKPVXgX8MNn0ZeFu+2hrU927gG/mqIw+2A93ufsjdp4D7iL22eePu33P36eDuo8CGfNa3gEx+/x3E3jsQey9db3m+CK+7n3D3nwe3R4D9wPp81hmiHcBXPOZRYIWZrS1g/dcDL7p7LieQZs3dfwT0J21OfA+ly6G3Ag+5e7+7DwAPATfl0paSDv95rAeOJtzv4dz/HCuBwYSgSVUmTL8CnHL3g2ked+B7ZvaEmd2Zx3Ykuiv46v3FNF81M3kd8+k3ifUSUynE65XJ7z9XJngvDRF7bxVEMMx0ObAnxcOvM7OnzOw7ZnZJgZq00N9lqd9Tt5G+A7YUrxfAanc/AbEPdmBVijKhv25FfwF3M/s+sCbFQ5909wfSPS3FtuRpTZmUyUiGbbyd+Xv917r7cTNbBTxkZs8HvYSszdcu4G+BTxP7nT9NbEjqN5N3keK5OU8Py+T1MrNPAtPA19PsJvTXK1VTU2zL2/toscysEfg/wO+6+3DSwz8nNrRxNjie88/A1gI0a6G/y1K+XhHgVuATKR5eqtcrU6G/bkUf/u5+QxZP6wE2JtzfABxPKnOG2FfOqqDHlqpMKG00syrgHcCV8+zjePDvaTP7J2JDDjmFWaavnZl9Dvh2iocyeR1Db1dwMOvXges9GPBMsY/QX68UMvn942V6gr9zC+d+rQ+dmVUTC/6vu/u3kh9P/DBw991m9jdm1u7ueV3HJoO/S17eUxm6Gfi5u59KfmCpXq/AKTNb6+4ngiGw0ynK9BA7LhG3gdixzqwt12GfXcBtwUyMLcQ+wR9LLBCEyg+Bdwab7gDSfZPI1Q3A8+7ek+pBM2sws6b4bWIHPfelKhuWpHHWt6ep73Fgq8VmRUWIfWXeled23QT8Z+BWdx9LU6ZQr1cmv/8uYu8diL2XfpDuAysswTGFLwD73f1/pCmzJn7swcy2E/u/3pfndmXyd9kFfCCY9XMNMBQf8iiAtN++l+L1SpD4HkqXQw8CbzGz1mCI9i3Btuzl++h2Pn+IhVYPMAmcAh5MeOyTxGZqHABuTti+G1gX3D6P2IdCN/APQE2e2vkl4CNJ29YBuxPa8VTw8yyx4Y98v3ZfBZ4Bng7efGuT2xXcv4XYbJIXC9SubmJjm08GP59NblchX69Uvz/wKWIfTgC1wXunO3gvnVeA1+gNxL7yP53wOt0CfCT+PgPuCl6bp4gdOH99AdqV8u+S1C4D7g1ez2dImKWX57bVEwvzloRtBX+9iH34nACiQXZ9iNgxooeBg8G/bUHZLuDzCc/9zeB91g18MNe26AxfEZEytFyHfUREZB4KfxGRMqTwFxEpQwp/EZEypPAXESlDCn8RkTKk8BcRKUMKfxGRMvT/AQEvVR8WE9DIAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(np.reshape(check_values,(784,)),np.reshape(edo,(784,)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "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": "3.6.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}