2.RedesNeuronalesDensamenteConectas.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 2 Redes neuronales densamente conectadas\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'2.1.3'"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import keras\n",
    "keras.__version__"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Precarga de los datos en Keras"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.datasets import mnist\n",
    "\n",
    "# obtenemos los datos para train y test \n",
    "(x_train, y_train), (x_test, y_test) = mnist.load_data()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "3\n"
     ]
    }
   ],
   "source": [
    "print(x_train.ndim) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(60000, 28, 28)\n"
     ]
    }
   ],
   "source": [
    "print(x_train.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "uint8\n"
     ]
    }
   ],
   "source": [
    "print(x_train.dtype)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "60000"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len (y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "1\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x10c323668>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "plt.imshow(x_train[8], cmap=plt.cm.binary)\n",
    "print(y_train[8])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[  0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0]\n",
      " [  0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0]\n",
      " [  0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0]\n",
      " [  0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0]\n",
      " [  0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0]\n",
      " [  0   0   0   0   0   0   0   0   0   0   0   0   5  63 197   0   0   0   0   0   0   0   0   0   0   0   0   0]\n",
      " [  0   0   0   0   0   0   0   0   0   0   0   0  20 254 230  24   0   0   0   0   0   0   0   0   0   0   0   0]\n",
      " [  0   0   0   0   0   0   0   0   0   0   0   0  20 254 254  48   0   0   0   0   0   0   0   0   0   0   0   0]\n",
      " [  0   0   0   0   0   0   0   0   0   0   0   0  20 254 255  48   0   0   0   0   0   0   0   0   0   0   0   0]\n",
      " [  0   0   0   0   0   0   0   0   0   0   0   0  20 254 254  57   0   0   0   0   0   0   0   0   0   0   0   0]\n",
      " [  0   0   0   0   0   0   0   0   0   0   0   0  20 254 254 108   0   0   0   0   0   0   0   0   0   0   0   0]\n",
      " [  0   0   0   0   0   0   0   0   0   0   0   0  16 239 254 143   0   0   0   0   0   0   0   0   0   0   0   0]\n",
      " [  0   0   0   0   0   0   0   0   0   0   0   0   0 178 254 143   0   0   0   0   0   0   0   0   0   0   0   0]\n",
      " [  0   0   0   0   0   0   0   0   0   0   0   0   0 178 254 143   0   0   0   0   0   0   0   0   0   0   0   0]\n",
      " [  0   0   0   0   0   0   0   0   0   0   0   0   0 178 254 162   0   0   0   0   0   0   0   0   0   0   0   0]\n",
      " [  0   0   0   0   0   0   0   0   0   0   0   0   0 178 254 240   0   0   0   0   0   0   0   0   0   0   0   0]\n",
      " [  0   0   0   0   0   0   0   0   0   0   0   0   0 113 254 240   0   0   0   0   0   0   0   0   0   0   0   0]\n",
      " [  0   0   0   0   0   0   0   0   0   0   0   0   0  83 254 245  31   0   0   0   0   0   0   0   0   0   0   0]\n",
      " [  0   0   0   0   0   0   0   0   0   0   0   0   0  79 254 246  38   0   0   0   0   0   0   0   0   0   0   0]\n",
      " [  0   0   0   0   0   0   0   0   0   0   0   0   0   0 214 254 150   0   0   0   0   0   0   0   0   0   0   0]\n",
      " [  0   0   0   0   0   0   0   0   0   0   0   0   0   0 144 241   8   0   0   0   0   0   0   0   0   0   0   0]\n",
      " [  0   0   0   0   0   0   0   0   0   0   0   0   0   0 144 240   2   0   0   0   0   0   0   0   0   0   0   0]\n",
      " [  0   0   0   0   0   0   0   0   0   0   0   0   0   0 144 254  82   0   0   0   0   0   0   0   0   0   0   0]\n",
      " [  0   0   0   0   0   0   0   0   0   0   0   0   0   0 230 247  40   0   0   0   0   0   0   0   0   0   0   0]\n",
      " [  0   0   0   0   0   0   0   0   0   0   0   0   0   0 168 209  31   0   0   0   0   0   0   0   0   0   0   0]\n",
      " [  0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0]\n",
      " [  0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0]\n",
      " [  0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0   0]]\n"
     ]
    }
   ],
   "source": [
    "import numpy\n",
    "from numpy import linalg\n",
    "numpy.set_printoptions(precision=2, suppress=True, linewidth=120)\n",
    "print(numpy.matrix(x_train[8]))\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "4\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAD8CAYAAAC4nHJkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMS4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvNQv5yAAADdRJREFUeJzt3W+MVPW9x/HPV4XE2CZCWTcEsNsCudGQSJsJMSm5Vrk01jRiidmUB7gq6fqgRis8uEQlVx9o9Ia2aaKioBtorfwxrZEH5t56oUoasWE01AVs1WsWYUVYQmOtieEi3/tgD82Ke34zzJyZM8v3/Uo2O3O+85vzZfSzZ+acM+dn7i4A8VxQdgMAykH4gaAIPxAU4QeCIvxAUIQfCIrwA0ERfiAowg8EdVE7VzZt2jTv6elp5yqBUIaGhnT8+HGr57FNhd/Mrpf0S0kXSnra3R9JPb6np0fVarWZVQJIqFQqdT+24bf9ZnahpMclfV/SlZKWmdmVjT4fgPZq5jP/Aknvufv77n5S0hZJS4ppC0CrNRP+GZIOjbl/OFv2BWbWb2ZVM6uOjIw0sToARWr53n53X+/uFXevdHV1tXp1AOrUTPiHJc0ac39mtgzABNBM+PdImmtm3zCzyZJ+JGl7MW0BaLWGD/W5+ykzu1PSf2v0UN+Au+8vrDMALdXUcX53f0nSSwX1AqCNOL0XCIrwA0ERfiAowg8ERfiBoAg/EBThB4Ii/EBQhB8IivADQRF+ICjCDwRF+IGgCD8QFOEHgiL8QFCEHwiK8ANBEX4gKMIPBEX4gaDaOkU3MFFcd911TY3fuXNnQZ20Dlt+ICjCDwRF+IGgCD8QFOEHgiL8QFCEHwiqqeP8ZjYk6RNJn0s65e6VIpoCWu2ee+5J1nfv3p2s33LLLUW2U4oiTvK51t2PF/A8ANqIt/1AUM2G3yX93szeMLP+IhoC0B7Nvu1f6O7DZnaZpJfN7C/uvmvsA7I/Cv2SdPnllze5OgBFaWrL7+7D2e9jkl6QtGCcx6x394q7V7q6uppZHYACNRx+M7vEzL565rak70naV1RjAFqrmbf93ZJeMLMzz/Ocu/9XIV0BaLmGw+/u70u6qsBegEKtXr06t/bkk08mx06aNClZX7RoUUM9dRIO9QFBEX4gKMIPBEX4gaAIPxAU4QeC4tLdOG+9/vrrubWTJ08mxy5cuDBZ7+3tbainTsKWHwiK8ANBEX4gKMIPBEX4gaAIPxAU4QeC4jj/eW7Xrl3J+kMPPZSsb968OVmfOnXqOfdUlFq9DQ4O5tbmzJmTHLt27dqGeppI2PIDQRF+ICjCDwRF+IGgCD8QFOEHgiL8QFAc5z/P9fenp1B85513kvUDBw4k67W+995Ktc5ROHHiRG7t6aefTo696qrz/6r0bPmBoAg/EBThB4Ii/EBQhB8IivADQRF+IKiax/nNbEDSDyQdc/d52bKpkrZK6pE0JKnX3f/WujbRqIsvvjhZN7Nk/bPPPiuynXOyd+/eZP2DDz5I1lP/tjL/XZ2ini3/RknXn7VstaQd7j5X0o7sPoAJpGb43X2XpLNPlVoiaVN2e5OkmwruC0CLNfqZv9vdj2S3P5LUXVA/ANqk6R1+7u6SPK9uZv1mVjWz6sjISLOrA1CQRsN/1MymS1L2+1jeA919vbtX3L3S1dXV4OoAFK3R8G+X1Jfd7pP0YjHtAGiXmuE3s82Sdkv6FzM7bGYrJD0iabGZvSvp37L7ACaQmsf53X1ZTmlRwb2gQWvWrMmt7du3Lzn2iiuuSNZb+b32Tz/9NFl/9NFHmxp/9dVX59Zuvvnm5NgIOMMPCIrwA0ERfiAowg8ERfiBoAg/EBSX7p4ADh06lKxv2LAht3bRRen/xI8//niy3sqzMleuXJmsb9u2LVmfMWNGsv7aa6+dc0+RsOUHgiL8QFCEHwiK8ANBEX4gKMIPBEX4gaA4zt8BBgcHk/WlS5cm66nLo911113Jsddcc02y3qy1a9fm1jZu3NjUc993331NjY+OLT8QFOEHgiL8QFCEHwiK8ANBEX4gKMIPBMVx/gKcOnUqWX/22WeT9dtvvz1ZH50RLV9qKurdu3cnxz788MPJ+qpVq5L1EyfOnsP1i55//vncWq1/V19fX7J+xx13JOtIY8sPBEX4gaAIPxAU4QeCIvxAUIQfCIrwA0HVPM5vZgOSfiDpmLvPy5Y9IOnHks58kfxed3+pVU12ui1btiTrK1asSNZTx+nrMXfu3Nzanj17kmNr1bdv356sDw8PJ+sffvhhbu2yyy5Ljh0YGEjW0Zx6tvwbJV0/zvJfuPv87Cds8IGJqmb43X2XpPRpXAAmnGY+899pZm+Z2YCZTSmsIwBt0Wj410maLWm+pCOSfpb3QDPrN7OqmVVT15oD0F4Nhd/dj7r75+5+WtIGSQsSj13v7hV3r7Ry0kcA56ah8JvZ9DF3fyhpXzHtAGiXeg71bZb0XUnTzOywpP+Q9F0zmy/JJQ1J4ruVwARTM/zuvmycxc+0oJeOtnXr1tzabbfdlhw7efLkZP3SSy9N1p977rlkfcqU/P2tK1euTI599dVXk/Va5wE0c62B48ePJ8fOmjUrWX/llVeS9dmzZyfr0XGGHxAU4QeCIvxAUIQfCIrwA0ERfiAoLt1dp6eeeiq3VuuQ1P3335+s17p0dzMee+yxZL2/vz9Zr3Xp72acPn06Wb/22muTdQ7lNYctPxAU4QeCIvxAUIQfCIrwA0ERfiAowg8ExXH+Oi1ZsiS3tnTp0uTYWucBtFKtr83u37+/qeevddnyefPmNfzcM2fObHgsamPLDwRF+IGgCD8QFOEHgiL8QFCEHwiK8ANBcZy/TnfffXfZLeT6+OOPc2vbtm1reKwkzZkzJ1nv7e1N1tG52PIDQRF+ICjCDwRF+IGgCD8QFOEHgiL8QFA1j/Ob2SxJv5LULcklrXf3X5rZVElbJfVIGpLU6+5/a12ryPPEE0/k1tatW5cc293dnazv3LmzoZ7Q+erZ8p+StMrdr5R0taSfmNmVklZL2uHucyXtyO4DmCBqht/dj7j7m9ntTyS9LWmGpCWSNmUP2yTpplY1CaB45/SZ38x6JH1L0p8kdbv7kaz0kUY/FgCYIOoOv5l9RdJvJf3U3f8+tuburtH9AeON6zezqplVR0ZGmmoWQHHqCr+ZTdJo8H/j7r/LFh81s+lZfbqkY+ONdff17l5x90pXV1cRPQMoQM3wm5lJekbS2+7+8zGl7ZL6stt9kl4svj0ArVLPV3q/I2m5pEEz25stu1fSI5K2mdkKSQcl8d3OFjl48GCyvmHDhtzaBRek/77XmqKby2efv2qG393/KMlyyouKbQdAu3CGHxAU4QeCIvxAUIQfCIrwA0ERfiAoLt09ASxevDhZT50HsHz58uTYBx98sKGeMPGx5QeCIvxAUIQfCIrwA0ERfiAowg8ERfiBoDjOPwHceuutyfqaNWtyazfeeGPB3eB8wZYfCIrwA0ERfiAowg8ERfiBoAg/EBThB4Ky0Zm22qNSqXi1Wm3b+oBoKpWKqtVq3qX2v4AtPxAU4QeCIvxAUIQfCIrwA0ERfiAowg8EVTP8ZjbLzP5gZgfMbL+Z3Z0tf8DMhs1sb/ZzQ+vbBVCUei7mcUrSKnd/08y+KukNM3s5q/3C3de2rj0ArVIz/O5+RNKR7PYnZva2pBmtbgxAa53TZ34z65H0LUl/yhbdaWZvmdmAmU3JGdNvZlUzq46MjDTVLIDi1B1+M/uKpN9K+qm7/13SOkmzJc3X6DuDn403zt3Xu3vF3StdXV0FtAygCHWF38wmaTT4v3H330mSux9198/d/bSkDZIWtK5NAEWrZ2+/SXpG0tvu/vMxy6ePedgPJe0rvj0ArVLP3v7vSFouadDM9mbL7pW0zMzmS3JJQ5LuaEmHAFqinr39f5Q03veDXyq+HQDtwhl+QFCEHwiK8ANBEX4gKMIPBEX4gaAIPxAU4QeCIvxAUIQfCIrwA0ERfiAowg8ERfiBoNo6RbeZjUg6OGbRNEnH29bAuenU3jq1L4neGlVkb19397qul9fW8H9p5WZVd6+U1kBCp/bWqX1J9NaosnrjbT8QFOEHgio7/OtLXn9Kp/bWqX1J9NaoUnor9TM/gPKUveUHUJJSwm9m15vZX83sPTNbXUYPecxsyMwGs5mHqyX3MmBmx8xs35hlU83sZTN7N/s97jRpJfXWETM3J2aWLvW167QZr9v+tt/MLpT0jqTFkg5L2iNpmbsfaGsjOcxsSFLF3Us/Jmxm/yrpH5J+5e7zsmX/KemEuz+S/eGc4u7/3iG9PSDpH2XP3JxNKDN97MzSkm6SdKtKfO0SffWqhNetjC3/Aknvufv77n5S0hZJS0roo+O5+y5JJ85avETSpuz2Jo3+z9N2Ob11BHc/4u5vZrc/kXRmZulSX7tEX6UoI/wzJB0ac/+wOmvKb5f0ezN7w8z6y25mHN3ZtOmS9JGk7jKbGUfNmZvb6ayZpTvmtWtkxuuiscPvyxa6+7clfV/ST7K3tx3JRz+zddLhmrpmbm6XcWaW/qcyX7tGZ7wuWhnhH5Y0a8z9mdmyjuDuw9nvY5JeUOfNPnz0zCSp2e9jJffzT500c/N4M0urA167Tprxuozw75E018y+YWaTJf1I0vYS+vgSM7sk2xEjM7tE0vfUebMPb5fUl93uk/Riib18QafM3Jw3s7RKfu06bsZrd2/7j6QbNLrH/38l3VdGDzl9fVPSn7Of/WX3JmmzRt8G/p9G942skPQ1STskvSvpfyRN7aDefi1pUNJbGg3a9JJ6W6jRt/RvSdqb/dxQ9muX6KuU140z/ICg2OEHBEX4gaAIPxAU4QeCIvxAUIQfCIrwA0ERfiCo/webFEk6bu28lwAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x10c3bd4a8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.imshow(x_train[9], cmap=plt.cm.binary)\n",
    "print(y_train[9])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "3\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x10c3316d8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.imshow(x_train[10], cmap=plt.cm.binary)\n",
    "print(y_train[10])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "6\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAP8AAAD8CAYAAAC4nHJkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMS4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvNQv5yAAADnlJREFUeJzt3X+MVPW5x/HPo5Y/tJio7CUI6vZW8EdIpDAhBM1NDZZYbRbQxNQYgwneJVqTW+GPGiupUWKMuaVBc4MuP1J60ys0aVUi5lovNrH1B2Hc5fqj3ivULBZcYAkaJCb2os/9Y4/NKjvfM8ycmTPr834lm505z5w5DwMfzpz5zjlfc3cBiOe0shsAUA7CDwRF+IGgCD8QFOEHgiL8QFCEHwiK8ANBEX4gqDPaubFJkyZ5d3d3OzcJhDI4OKgjR45YPY9tKvxmdq2ktZJOl7TB3R9OPb67u1vVarWZTQJIqFQqdT+24bf9Zna6pH+T9H1Jl0u62cwub/T5ALRXM8f8cyXtdff33P1vkrZIWlRMWwBarZnwT5X011H392fLvsTMes2sambV4eHhJjYHoEgt/7Tf3fvcveLula6urlZvDkCdmgn/AUkXjLo/LVsGYBxoJvy7JE03s2+Z2QRJP5S0rZi2ALRaw0N97n7CzO6S9LxGhvo2ufvbhXUGoKWaGud39+ckPVdQLwDaiK/3AkERfiAowg8ERfiBoAg/EBThB4Ii/EBQhB8IivADQRF+ICjCDwRF+IGgCD8QVFsv3Y3GfPrpp8n6/Pnza9YGBgaS6/b09CTrTz/9dLKO8Ys9PxAU4QeCIvxAUIQfCIrwA0ERfiAowg8ExTh/B8gbx7/77ruT9d27d9esmaVna54zZ06yjq8v9vxAUIQfCIrwA0ERfiAowg8ERfiBoAg/EFRT4/xmNijpY0mfSTrh7pUimorm0UcfTdafeOKJZH3BggU1aw888EBy3Xnz5iXr+Poq4ks+V7v7kQKeB0Ab8bYfCKrZ8Luk35vZ62bWW0RDANqj2bf9V7n7ATP7B0kvmNn/uPtLox+Q/afQK0kXXnhhk5sDUJSm9vzufiD7fVjSU5LmjvGYPnevuHulq6urmc0BKFDD4Tezs8xs4he3JS2U9FZRjQForWbe9k+W9FR2yugZkv7D3f+zkK4AtFzD4Xf39yRdUWAvYQ0NDTW1/jXXXFOzxjg+amGoDwiK8ANBEX4gKMIPBEX4gaAIPxAUl+7uAMePH0/WJ0yYkKynhvqAWtjzA0ERfiAowg8ERfiBoAg/EBThB4Ii/EBQjPO3wQcffJCsb9iwIVmfP39+sj579uxT7glgzw8ERfiBoAg/EBThB4Ii/EBQhB8IivADQTHO3warV68uu4Vx6dVXX03W9+/f3/BzX3FF+qrzM2bMaPi5xwv2/EBQhB8IivADQRF+ICjCDwRF+IGgCD8QVO44v5ltkvQDSYfdfWa27FxJWyV1SxqUdJO7f9i6Nse37du3N7X+7bffXlAn7XfHHXfUrOW9Lh9+mP4n9cknnzTUkySdffbZyfqKFSuS9VWrVjW87U5Rz57/l5Ku/cqyeyTtcPfpknZk9wGMI7nhd/eXJB39yuJFkjZntzdLWlxwXwBarNFj/snuPpTdPihpckH9AGiTpj/wc3eX5LXqZtZrZlUzqw4PDze7OQAFaTT8h8xsiiRlvw/XeqC797l7xd0rXV1dDW4OQNEaDf82SUuz20slPVNMOwDaJTf8ZvakpFclXWJm+81smaSHJX3PzPZIuia7D2AcsZFD9vaoVCperVbbtr12yRtvvvjii5P1M85If93i/fffP+We6nXixIlkvb+/P1lfvDg90HPw4MGatbx/e3mHiVdeeWWynuo97zWdOnVqsv7yyy8n6xdddFGy3iqVSkXVatXqeSzf8AOCIvxAUIQfCIrwA0ERfiAowg8ExaW7C5A3xfahQ4eS9eXLlxfZzpfkTQ/e19eXrD/44INNbT81ZHbrrbcm173zzjuT9WnTpjXUkyT19PQk63mnGw8NDSXrZQ31nQr2/EBQhB8IivADQRF+ICjCDwRF+IGgCD8QFOP8BRgYGGhq/enTpxfUycnypgd//PHHk3Wz9NmhCxYsSNbXrFlTszZz5szkuq2Ud5p1BOz5gaAIPxAU4QeCIvxAUIQfCIrwA0ERfiAoxvkLkHfOfKu9++67NWtbtmxp6rl7e3uT9bVr1ybrEyZMaGr7ZZkzZ06yPnv27DZ10jrs+YGgCD8QFOEHgiL8QFCEHwiK8ANBEX4gqNxxfjPbJOkHkg67+8xs2f2S/lnScPawe939uVY12emOHTuWrOdNRd3sNOmPPfZYzdpHH32UXPeWW25J1tetW9dQT53u+PHjyXretOnj9fsLo9Wz5/+lpGvHWP4Ld5+V/YQNPjBe5Ybf3V+SdLQNvQBoo2aO+e8yszfMbJOZnVNYRwDaotHwr5P0bUmzJA1J+nmtB5pZr5lVzaw6PDxc62EA2qyh8Lv7IXf/zN0/l7Re0tzEY/vcveLula6urkb7BFCwhsJvZlNG3V0i6a1i2gHQLvUM9T0p6buSJpnZfkk/k/RdM5slySUNSmrdHNMAWiI3/O5+8xiLN7agl3Er79r2zdbzpK4nkPfcZV+LoJVSf7YNGzYk173xxhuLbqfj8A0/ICjCDwRF+IGgCD8QFOEHgiL8QFBcuvtroK+vr2btlVdeSa6bV3/ooYeS9eXL01/xOO+885L1Vrrhhhtq1s4888zkuitXriy6nY7Dnh8IivADQRF+ICjCDwRF+IGgCD8QFOEHgmKcv06p00OHhoba2MnJUmPp/f39yXV7enqS9VWrViXrzz//fLL+7LPP1qxNnDix4XUlafXq1cn6wMBAzdp9992XXHfevHnJ+tcBe34gKMIPBEX4gaAIPxAU4QeCIvxAUIQfCIpx/jqdf/75NWszZsxIrrtv375k/cUXX0zW886ZT52bPmXKlJo1Sdq1a1eynjfWftlllyXrqSnC886Zz7u8dt45+amx/LzvL0TAnh8IivADQRF+ICjCDwRF+IGgCD8QFOEHgsod5zezCyT9StJkSS6pz93Xmtm5krZK6pY0KOkmd/+wda12ro0b0zOWX3/99cn69u3bk/WFCxcm6ytWrKhZyxvnz7Nz585kPe+6/qn13T257iWXXNLUtpcsWZKsR1fPnv+EpJXufrmkeZJ+ZGaXS7pH0g53ny5pR3YfwDiRG353H3L3/uz2x5LekTRV0iJJm7OHbZa0uFVNAijeKR3zm1m3pO9I2ilpsrt/cf2qgxo5LAAwTtQdfjP7pqTfSvqxux8bXfORg7cxD+DMrNfMqmZWHR4ebqpZAMWpK/xm9g2NBP/X7v67bPEhM5uS1adIOjzWuu7e5+4Vd690dXUV0TOAAuSG38xM0kZJ77j7mlGlbZKWZreXSnqm+PYAtIrlDbeY2VWS/ijpTUmfZ4vv1chx/28kXShpn0aG+o6mnqtSqXi1Wm2253En79LeV199dbK+Z8+eItv5kjr+/lu27dtuuy1Zf+SRR5L1Mqf/7lSVSkXVarWuv7TccX53/5OkWk+24FQaA9A5+IYfEBThB4Ii/EBQhB8IivADQRF+ICgu3d0GeafVvvbaa8n61q1bk/W9e/fWrK1fvz657rJly5L1005rbv+Qev5LL720qedGc9jzA0ERfiAowg8ERfiBoAg/EBThB4Ii/EBQuefzFynq+fxAu5zK+fzs+YGgCD8QFOEHgiL8QFCEHwiK8ANBEX4gKMIPBEX4gaAIPxAU4QeCIvxAUIQfCIrwA0ERfiCo3PCb2QVm9gcz+7OZvW1m/5Itv9/MDpjZ7uznuta3C6Ao9UzacULSSnfvN7OJkl43sxey2i/c/V9b1x6AVskNv7sPSRrKbn9sZu9ImtrqxgC01ikd85tZt6TvSNqZLbrLzN4ws01mdk6NdXrNrGpm1eHh4aaaBVCcusNvZt+U9FtJP3b3Y5LWSfq2pFkaeWfw87HWc/c+d6+4e6Wrq6uAlgEUoa7wm9k3NBL8X7v77yTJ3Q+5+2fu/rmk9ZLmtq5NAEWr59N+k7RR0jvuvmbU8tFTzy6R9Fbx7QFolXo+7b9S0q2S3jSz3dmyeyXdbGazJLmkQUnLW9IhgJao59P+P0ka6zrgzxXfDoB24Rt+QFCEHwiK8ANBEX4gKMIPBEX4gaAIPxAU4QeCIvxAUIQfCIrwA0ERfiAowg8ERfiBoMzd27cxs2FJ+0YtmiTpSNsaODWd2lun9iXRW6OK7O0id6/renltDf9JGzerunultAYSOrW3Tu1LordGldUbb/uBoAg/EFTZ4e8refspndpbp/Yl0VujSumt1GN+AOUpe88PoCSlhN/MrjWz/zWzvWZ2Txk91GJmg2b2ZjbzcLXkXjaZ2WEze2vUsnPN7AUz25P9HnOatJJ664iZmxMzS5f62nXajNdtf9tvZqdLelfS9yTtl7RL0s3u/ue2NlKDmQ1Kqrh76WPCZvZPko5L+pW7z8yWPSLpqLs/nP3HeY67/6RDertf0vGyZ27OJpSZMnpmaUmLJd2mEl+7RF83qYTXrYw9/1xJe939PXf/m6QtkhaV0EfHc/eXJB39yuJFkjZntzdr5B9P29XorSO4+5C792e3P5b0xczSpb52ib5KUUb4p0r666j7+9VZU367pN+b2etm1lt2M2OYnE2bLkkHJU0us5kx5M7c3E5fmVm6Y167Rma8Lhof+J3sKnefLen7kn6Uvb3tSD5yzNZJwzV1zdzcLmPMLP13Zb52jc54XbQywn9A0gWj7k/LlnUEdz+Q/T4s6Sl13uzDh76YJDX7fbjkfv6uk2ZuHmtmaXXAa9dJM16XEf5dkqab2bfMbIKkH0raVkIfJzGzs7IPYmRmZ0laqM6bfXibpKXZ7aWSnimxly/plJmba80srZJfu46b8drd2/4j6TqNfOL/F0k/LaOHGn39o6T/zn7eLrs3SU9q5G3g/2nks5Flks6TtEPSHkn/JencDurt3yW9KekNjQRtSkm9XaWRt/RvSNqd/VxX9muX6KuU141v+AFB8YEfEBThB4Ii/EBQhB8IivADQRF+ICjCDwRF+IGg/h92sok3vzEgmAAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x10c490e80>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.imshow(x_test[11], cmap=plt.cm.binary)\n",
    "print(y_test[11])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [],
   "source": [
    "x_train = x_train.astype('float32')\n",
    "x_test = x_test.astype('float32')\n",
    "\n",
    "x_train /= 255\n",
    "x_test /= 255\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.\n",
      "  0.   0.   0.   0.   0.  ]\n",
      " [0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.\n",
      "  0.   0.   0.   0.   0.  ]\n",
      " [0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.\n",
      "  0.   0.   0.   0.   0.  ]\n",
      " [0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.\n",
      "  0.   0.   0.   0.   0.  ]\n",
      " [0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.\n",
      "  0.   0.   0.   0.   0.  ]\n",
      " [0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.02 0.25 0.77 0.   0.   0.   0.   0.   0.   0.   0.\n",
      "  0.   0.   0.   0.   0.  ]\n",
      " [0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.08 1.   0.9  0.09 0.   0.   0.   0.   0.   0.   0.\n",
      "  0.   0.   0.   0.   0.  ]\n",
      " [0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.08 1.   1.   0.19 0.   0.   0.   0.   0.   0.   0.\n",
      "  0.   0.   0.   0.   0.  ]\n",
      " [0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.08 1.   1.   0.19 0.   0.   0.   0.   0.   0.   0.\n",
      "  0.   0.   0.   0.   0.  ]\n",
      " [0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.08 1.   1.   0.22 0.   0.   0.   0.   0.   0.   0.\n",
      "  0.   0.   0.   0.   0.  ]\n",
      " [0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.08 1.   1.   0.42 0.   0.   0.   0.   0.   0.   0.\n",
      "  0.   0.   0.   0.   0.  ]\n",
      " [0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.06 0.94 1.   0.56 0.   0.   0.   0.   0.   0.   0.\n",
      "  0.   0.   0.   0.   0.  ]\n",
      " [0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.7  1.   0.56 0.   0.   0.   0.   0.   0.   0.\n",
      "  0.   0.   0.   0.   0.  ]\n",
      " [0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.7  1.   0.56 0.   0.   0.   0.   0.   0.   0.\n",
      "  0.   0.   0.   0.   0.  ]\n",
      " [0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.7  1.   0.64 0.   0.   0.   0.   0.   0.   0.\n",
      "  0.   0.   0.   0.   0.  ]\n",
      " [0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.7  1.   0.94 0.   0.   0.   0.   0.   0.   0.\n",
      "  0.   0.   0.   0.   0.  ]\n",
      " [0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.44 1.   0.94 0.   0.   0.   0.   0.   0.   0.\n",
      "  0.   0.   0.   0.   0.  ]\n",
      " [0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.33 1.   0.96 0.12 0.   0.   0.   0.   0.   0.\n",
      "  0.   0.   0.   0.   0.  ]\n",
      " [0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.31 1.   0.96 0.15 0.   0.   0.   0.   0.   0.\n",
      "  0.   0.   0.   0.   0.  ]\n",
      " [0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.84 1.   0.59 0.   0.   0.   0.   0.   0.\n",
      "  0.   0.   0.   0.   0.  ]\n",
      " [0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.56 0.95 0.03 0.   0.   0.   0.   0.   0.\n",
      "  0.   0.   0.   0.   0.  ]\n",
      " [0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.56 0.94 0.01 0.   0.   0.   0.   0.   0.\n",
      "  0.   0.   0.   0.   0.  ]\n",
      " [0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.56 1.   0.32 0.   0.   0.   0.   0.   0.\n",
      "  0.   0.   0.   0.   0.  ]\n",
      " [0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.9  0.97 0.16 0.   0.   0.   0.   0.   0.\n",
      "  0.   0.   0.   0.   0.  ]\n",
      " [0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.66 0.82 0.12 0.   0.   0.   0.   0.   0.\n",
      "  0.   0.   0.   0.   0.  ]\n",
      " [0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.\n",
      "  0.   0.   0.   0.   0.  ]\n",
      " [0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.\n",
      "  0.   0.   0.   0.   0.  ]\n",
      " [0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.   0.\n",
      "  0.   0.   0.   0.   0.  ]]\n"
     ]
    }
   ],
   "source": [
    "print(numpy.matrix(x_train[8]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(60000, 784)\n",
      "(10000, 784)\n"
     ]
    }
   ],
   "source": [
    "\n",
    "x_train = x_train.reshape(60000, 784)\n",
    "x_test = x_test.reshape(10000, 784)\n",
    "\n",
    "\n",
    "print(x_train.shape)\n",
    "print(x_test.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.utils import to_categorical\n",
    "\n",
    "y_train = keras.utils.to_categorical(y_train, num_classes=10)\n",
    "y_test = keras.utils.to_categorical(y_test, num_classes=10)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Definición del modelo"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "dense_3 (Dense)              (None, 10)                7850      \n",
      "_________________________________________________________________\n",
      "dense_4 (Dense)              (None, 10)                110       \n",
      "=================================================================\n",
      "Total params: 7,960\n",
      "Trainable params: 7,960\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "from keras.models import Sequential\n",
    "from keras.layers import Dense\n",
    "from keras.optimizers import sgd\n",
    "\n",
    "model = Sequential()\n",
    "model.add(Dense(10, activation='sigmoid', input_shape=(784,)))\n",
    "model.add(Dense(10, activation='softmax'))\n",
    "\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Aprendizaje, entrenamiento y evaluación"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/10\n",
      "60000/60000 [==============================] - 2s 28us/step - loss: 2.0162 - acc: 0.4937\n",
      "Epoch 2/10\n",
      "60000/60000 [==============================] - 1s 24us/step - loss: 1.5334 - acc: 0.6890\n",
      "Epoch 3/10\n",
      "60000/60000 [==============================] - 1s 21us/step - loss: 1.1935 - acc: 0.7699\n",
      "Epoch 4/10\n",
      "60000/60000 [==============================] - 1s 21us/step - loss: 0.9736 - acc: 0.8175\n",
      "Epoch 5/10\n",
      "60000/60000 [==============================] - 1s 21us/step - loss: 0.8290 - acc: 0.8421\n",
      "Epoch 6/10\n",
      "60000/60000 [==============================] - 1s 21us/step - loss: 0.7286 - acc: 0.8558\n",
      "Epoch 7/10\n",
      "60000/60000 [==============================] - 1s 20us/step - loss: 0.6553 - acc: 0.8662\n",
      "Epoch 8/10\n",
      "60000/60000 [==============================] - 1s 21us/step - loss: 0.5999 - acc: 0.8735\n",
      "Epoch 9/10\n",
      "60000/60000 [==============================] - 1s 20us/step - loss: 0.5569 - acc: 0.8782\n",
      "Epoch 10/10\n",
      "60000/60000 [==============================] - 1s 21us/step - loss: 0.5229 - acc: 0.8822\n",
      "10000/10000 [==============================] - 0s 18us/step\n",
      "Test loss: 0.49781588258743287\n",
      "Test accuracy: 0.8895\n"
     ]
    }
   ],
   "source": [
    "batch_size = 50\n",
    "num_classes = 10\n",
    "epochs=10\n",
    "\n",
    "model.compile(loss='categorical_crossentropy',\n",
    "              optimizer='sgd',\n",
    "              metrics=['accuracy'])\n",
    "\n",
    "model.fit(x_train, y_train,\n",
    "          batch_size=batch_size,\n",
    "          epochs=epochs,\n",
    "          verbose=1\n",
    "          )\n",
    "\n",
    "test_loss, test_acc = model.evaluate(x_test, y_test)\n",
    "\n",
    "print('Test loss:', test_loss)\n",
    "print('Test accuracy:', test_acc)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Predicciones"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [],
   "source": [
    "predictions = model.predict(x_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1.0000001"
      ]
     },
     "execution_count": 38,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import numpy as np\n",
    "np.sum(predictions[11])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "6"
      ]
     },
     "execution_count": 39,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.argmax(predictions[11])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Look at confusion matrix \n",
    "#Note, this code is taken straight from the SKLEARN website, an nice way of viewing confusion matrix.\n",
    "def plot_confusion_matrix(cm, classes,\n",
    "                          normalize=False,\n",
    "                          title='Confusion matrix',\n",
    "                          cmap=plt.cm.Blues):\n",
    "    \"\"\"\n",
    "    This function prints and plots the confusion matrix.\n",
    "    Normalization can be applied by setting `normalize=True`.\n",
    "    \"\"\"\n",
    "    plt.imshow(cm, interpolation='nearest', cmap=cmap)\n",
    "    plt.title(title)\n",
    "    plt.colorbar()\n",
    "    tick_marks = np.arange(len(classes))\n",
    "    plt.xticks(tick_marks, classes, rotation=45)\n",
    "    plt.yticks(tick_marks, classes)\n",
    "\n",
    "    if normalize:\n",
    "        cm = cm.astype('float') / cm.sum(axis=1)[:, np.newaxis]\n",
    "\n",
    "    thresh = cm.max() / 2.\n",
    "    for i, j in itertools.product(range(cm.shape[0]), range(cm.shape[1])):\n",
    "        plt.text(j, i, cm[i, j],\n",
    "                 horizontalalignment=\"center\",\n",
    "                 color=\"white\" if cm[i, j] > thresh else \"black\")\n",
    "\n",
    "    plt.tight_layout()\n",
    "    plt.ylabel('Observación')\n",
    "    plt.xlabel('Predicción')\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAVEAAAEmCAYAAADbUaM7AAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMS4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvNQv5yAAAIABJREFUeJzsnXd8FcXXh5+TBAIISg0loZck1PQgvfemdKkCYv29KhYURQRFkaKCiIiVpmCn916kE5EqICAJNfSewrx/7E24QMotu5CEefjsh7tlzszunZw7O+V8RSmFRqPRaFzD434XQKPRaDIz2olqNBqNG2gnqtFoNG6gnahGo9G4gXaiGo1G4wbaiWo0Go0baCf6ACEiOUVkjohcEJGf3bDTTUQWm1m2+4WI1BaRffe7HJrMi+h5ohkPEXkCGAAEAJeAKGC4Umqtm3Z7AP8DaiilEtwuaAZHRBRQXil14H6XRZN10S3RDIaIDAA+BT4ACgMlgAlAWxPMlwT+eRAcqCOIiNf9LoMmC6CU0lsG2YBHgMtAxzSu8cZwssds26eAt+1cPSAaeAU4BRwHnrSdGwrEAfG2PPoC7wLT7GyXAhTgZdvvDfyL0Ro+BHSzO77WLl0NYDNwwfZ/DbtzK4H3gHU2O4uBgqncW1L5X7crfzugBfAPcBYYZHd9BPAncN527Xggu+3catu9XLHdb2c7+wOBE8DUpGO2NGVteYTY9osBp4F697tu6C3jbrolmrF4FMgB/J7GNW8B1YEgoBqGI3nb7nwRDGfsi+EoPxeRfEqpIRit25lKqdxKqW/SKoiIPASMA5orpfJgOMqoFK7LD8yzXVsA+BiYJyIF7C57AngS8AGyA6+mkXURjGfgC7wDfAV0B0KB2sBgESltuzYReBkoiPHsGgLPASil6tiuqWa735l29vNjtMr722eslDqI4WCniUgu4DtgslJqZRrl1TzgaCeasSgAxKq0X7e7AcOUUqeUUqcxWpg97M7H287HK6XmY7TC/F0sz02gsojkVEodV0rtSuGalsB+pdRUpVSCUupHYC/Q2u6a75RS/yilrgE/YfwApEY8Rv9vPDADw0GOVUpdsuW/G+PHA6XUVqXUBlu+h4EvgboO3NMQpdQNW3luQyn1FXAA2AgUxfjR0mhSRTvRjMUZoGA6fXXFgCN2+0dsx5Jt3OGErwK5nS2IUuoKxivwM8BxEZknIgEOlCepTL52+yecKM8ZpVSi7XOSkztpd/5aUnoRqSAic0XkhIhcxGhpF0zDNsBppdT1dK75CqgMfKaUupHOtZoHHO1EMxZ/Ajcw+gFT4xjGq2gSJWzHXOEKkMtuv4j9SaXUIqVUY4wW2V4M55JeeZLKFONimZzhC4xylVdKPQwMAiSdNGlORxGR3Bj9zN8A79q6KzSaVNFONAOhlLqA0Q/4uYi0E5FcIpJNRJqLyEjbZT8Cb4tIIREpaLt+motZRgF1RKSEiDwCvJl0QkQKi0hbW9/oDYxugZsp2JgPVBCRJ0TES0Q6AxWBuS6WyRnyABeBy7ZW8rN3nD8JlHHS5lhgi1KqH0Zf70S3S6nJ0mgnmsFQSo3BmCP6NsbI8FHgBeAP2yXvA1uAHcDfwDbbMVfyWgLMtNnayu2Oz8NWjmMYI9Z1udtJoZQ6A7TCmBFwBmNkvZVSKtaVMjnJqxiDVpcwWskz7zj/LjBZRM6LSKf0jIlIW6AZt+5zABAiIt1MK7Emy6En22s0Go0b6JaoRqPRuIF2ohqNRuMG2olqNBqNG2gnqtFoNG6QoQIwiHce5fFQenOlXSOodIH0L3IRq4fm0pv4mFHJrEOWmfV5W8mRI4eJjY019dF4PlxSqYS7Fo2lirp2epFSqpmZZTCDDOVEPR4qSM5GQyyxve6H3pbYBSOIi5WIZM4/65s3rXsuVj5xT4/M+bytpGZkmOk2VcI1vP3TnXmWzPWoz61pYblJhnKiGo3mQUJAMn+PonaiGo3m/iBAJn3Lskc7UY1Gc//IAi3RDHsHzzUPZNPotmwe05bnWlS87dz/WlXi8k+9KZDH+7bjIWULcP7HnrSLvDMehuMsXrSQqpX8qRRQjlEjR7hsJyUCypcmPLgqkWHB1Kwebprd69evU+vRCCJCqhFSrRLvDTW3X9nsZ/JM/z6U9CtMWHCV5GOD3niN4CqBRIRWo0vHxzl//rxLtp/t34dSfoUJt7OdxLhPxpDb24PYWPdXpD7drw8livkQGlTZbVv32r6Vddw5BDw8Hd8yKBnSiVYsnpfeDStQd9Bcqr82m+YhfpQpnAcA3wK5aFi1GP+dvnxbGg8R3usWxrK/XA1oBImJibz0f88za84Ctu/Yzc8zfmTP7t1u3cudLFiynI1btrNuw2bTbHp7e7NwyXI2bfuLjVuiWLxoIRs3bDDFthXPpHuP3vwxZ8Ftxxo0bMzm7X+zaetflCtfntEjP3TJdrcUbANEHz3KsqVLKF6ihEt276RHr97MmrvQFFv30v69qONOIeL4lkHJkE7U3/cRNh84zbW4RBJvKtbuOUEbW+vyo14RvD19C3cOiD/TPJBZG49w+mJ6oSJTZ/OmTZQtW47SZcqQPXt2Onbuwtw5s9y5lXuCiJA7txGiMz4+noT4eNNG9K14JrVq1yF/vtsjzDVq3AQvL6N3KSKyOjExrkXSq1W7Dvny3R29buBrA3j/w49Mey61atchf37rouRZZT9D1XHBeJ13dMugZMiS7T56nhoBhcmf25uc2T1pEuyHX4GHaBlWnGNnr7LzyLnbri+aLxdtIkrw1eK9buV77FgMfn7Fk/d9ff1c/mNOCRGhdYum1IgM45uvJ5lmF4wWRmRoECWK+dCgUWMiIiNNsWv1M0mJKd9/R5Om5k0HnDt7FsWKFaNK1Wqm2cys3I/vM3WcaIVm4JaopQNLItIMIz6jJ/C1UsqhDph9MRf4ZNZOZr3dmKvXE/j78Fm8s3nw6mNVafv+3XLnI3tHMHj61rtapxmNpSvW4Ovry6lTp2jdvAn+/gHUql0n/YQO4OnpycatUZw/f57OHR5j186dVKpsTX+dlYwcMRwvLy+6dDUn+tzVq1cZPfJDZs1bZIo9jclk4Bamo1jmREXEE/gcaIyhsLhZRGYrpRzqgJmyYj9TVuwHYEjXEE6fv0ar8BL8OcpQDvYtkIu1H7Wm7pvzCC5bgO9fNKR1CjzsTdNgXxJuKuZu/s+pMhcr5kt09NHk/ZiYaHx9fdNI4RxJtnx8fGjdth1bNm8yzYkmkTdvXurWq8/ixQtNcaJWPxN7pk75ngXz5zFv4VLTXrv//fcghw8f4tFwQ9YpJjqaWtVDWbV2I4WLFEknddbjXn6fDpGBW5iOYmVLNAI4oJT6F0BEZmBopzvkRAs9nIPTF6/jV+Ah2kaUpP5b85iwYE/y+V3jO1DnzTmcuXSDyi/8mnx84nO1WLj1qNMOFCAsPJwDB/Zz+NAhivn68vPMGXw/9Qen7aTElStXuHnzJnny5OHKlSssW7qEN98abIrt06dPky1bNvLmzcu1a9dYtnQJr7w20BTbVj4TexYvWsinY0axcOlKcuXKlX4CB6lcuQqHo29JNFWsUJrV6zdTsGCGXPxiOffq+3QMPdk+PXwxorInEQ3c1VEnIv2xSddKrlvr26e/Up/8ebyJT7jJgG82cOFqnIVFNfDy8uKTseNp3bIpiYmJ9Ordh4qVKpli+9TJk3Tp+DgACQkJdOrS1bR+vxPHj/NUn14kJiZyU92kfYdOtGjZyhTbVjyTXj2eYM3qlZyJjaV8meK8PfhdRo8cwY24G7Ru0QSAiIhIxn3uvDJHbzvbFcoU563B79Lryb5ulTclenbvyppVK4mNjaVsKT8GvzOU3n3My8cq+1bWcafJIpPtLYtsLyIdgGY2rRpEpAcQqZR6IbU0nvlLK6vWzsfqtfP3HL12PutQMzKMrVu3mPpgPPIUU97B/R2+/vqaoVuVUuYv4ncTK1uiMUBxu30/7o0CpEajyRQIeGbcSfSOYmWHxGagvIiUFpHsQBdgtoX5aTSazEQWmSdqWUtUKZUgIi8AizCmOH2rlNplVX4ajSYTkkm7quyxdJ6oUmo+hi65RqPR3EHWGJ3P/Heg0WgyLyauWBKRb0XklIjstDuWX0SWiMh+2//5bMdFRMaJyAER2SEiIXZpetmu3y8ivdLLVztRjUZz/zC3T/R74M55g28Ay5RS5YFltn2A5kB529Yf+AIMpwsMwZiOGQEMSXK8qaGdqEajuT840wp1oCWqlFoNnL3jcFtgsu3zZKCd3fEpymADkFdEigJNgSVKqbNKqXPAEu52zLehgzJrNJr7h3N9ogVFZIvd/iSlVHqRfAorpY7bPp8ACts+p7QYyDeN46minahGo7l/ODc6H+vOZHullBIR09dpZCgnGlS6gGWqnPnCU10o5TbnNo+3zHZmxkOv/EmRhMSbltn28sxMPXT3ZHT+pIgUVUodt72un7IdT20xUAxQ747jK9PKIDM9cY1Gk5UQ7oU8yGwgaYS9FzDL7nhP2yh9deCC7bV/EdBERPLZBpSa2I6lSoZqiWo0mgcJc1uiIvIjRiuyoIhEY4yyjwB+EpG+wBEgSeh+PtACOABcBZ4EUEqdFZH3MFZcAgxTSt05WHUb2olqNJr7h4krlpRSXVM51TCFaxXwfCp2vgW+dTRf7UQ1Gs39Q69Yuj+4Ivk6cUg3jiz7kC0/D0o+9nijYLb+8hZXto4jpOLtKpCv9mnCzllD+Ov3wTR6NDD5eOMagfz1+2B2zhrCq082drrsiYmJVA8L5vG25sT7vBe2rZYHtkrC9+jRozRtVJ/gqhUJqVaJ8ePGZljb0UeP0qJJQ8KCKhMeXIUJ48cB8NabrxNStSLVw4Lo2sl1KWl7Mo5kMllCYynTOVFXJV+nztlA2+c/v+3YroPH6PLKV6zddvC24wFlitCxaQghHYbT5vkJjH2zEx4egoeH8OkbnWj7wgSC279Px2ahBJRxTmJi/Lix+AcGpn+hC1hl20p5YCslfL28vBgxcgzbd+xm1doNfDnx8wxr28vLiw8+GsWWqJ0sX72eSRMnsHfPbho0aMSmbTvYsCWKcuUrMGaUe04vQ0kmi2SJKE4Zt2Sp4Krk67ptBzl74eptx/YdOsn+I6fuurZVvar8vGgbcfEJHDl2hoNHYwmvXIrwyqU4eDSWwzFniE9I5OdF22hVr6rDZY+Ojmbhgnk82aefw2kygm0r5YGtlPAtWrQowSHGkug8efIQEBDIsWPmhLQ123aRokUJCr5lzz8ggGMxMTS0k5IOj4jkWHS0W+XOUJLJoFui94N7IfnqW+gRok/ckmWOOXWOYj6PUMznEaJP2h0/eQ7fQo84bPe1V15i+Icj8fAw/7FbadtK7pWE75HDh4mK2k54hDlS0lbaPnL4MDuiogi7w97Uyd/R2E1JmYwlmWyoNji6ZVQs+4tLKaLKg8z8eXPxKeRDSGhoprKdFbh8+TJdO7Vn1JhPefjhhzO07cuXL9O9a0dGjP74NnujRnyAl5cXnU2Sks4IGBJLmd+JWjk6/z0wHphiptF7Ifkac/oCfkVuBW7x9cnHsVMXAPArbHe8cD5iTl9wyOaf69cxd+5sFi6cz43r17l48SJP9uzOd1OmuV1eK21bjdXfZ3x8PF07tadz1260e+xx0+xaYTs+Pp7uXTrQqcsTtG13y960Kd+zYME85i5Y4rYzyVCSySJIFljVZllLNJWIKm5jL/kaFxfHzzNn0LJVG1PzmLdyBx2bhpA9mxclixWgXIlCbN55mC27jlCuRCFKFitANi9POjYNYd7KHQ7ZfG/4hxw8HM2+A4eZMn0G9eo3MM3JWWnbaqz8PpVSPPNUX/wDAnnx5QGm2LTKtlKK55/uh39AIP978eXk40sWL+TTj0cz85c/TJGSvhd/P86gW6ImYC+ZXLxEiXSudl3ydfKHvakdWp6CeXNzYOF7vDdxPucuXOHjgR0pmC83v417hh37Ymjz/Ofs+fcEvy7ezvZf3yIh8SYvjfjJplypePmjn5gz4Xk8PYTJszaw598T7j6CDI+V8sBWSviuX7eOH6ZPpXLlKkSGBgEw9P0PaNa8RYaz/ef6dfz4wzQqVa5CjQhjgGnIsPd5fcBL3Lhxg7YtmwLG4NLY8V+4XO4MJZlM5lWytccyyWQAESkFzFVKOTTBMDQ0TK3buCX9C11AByDRZBQyYwASKySTPfOXVrmbDnP4+oszej5wkskajUaTOmLbMjnaiWo0mvuCkLH7Oh3FyilOPwJ/Av4iEm2LoqLRaDTJ6IGlNEgjoopGo9EAWWNgSb/OazSa+4Z2ohqNRuMqemBJo9FoXEeQTBfrISW0E9VoNPcN/Tqv0Wg07pD5fWjGcqIK61ZznN4wzhK7AKWe+8Uy2wAHPjM3cMa9wkr5XitX/VhNwk3rVgl6uSyKeR8Q3RLVaDQat9BOVKPRaNxAO1GNRqNxkayy7FM7UY1Gc//I/D40c2gspSYne/bsWdq0aEJQJX/atGjCuXPn0rF0N8/270vp4kWICLklOLfjryjq16lBjYgQ6tSIYMvmTU7Z7N+oPKvebczKIY35ol8E3l4e/PFaPZYObsTSwY2IGtmS7557FICm1Yqy/B3j+KJBDYgoV8ChPKyW2LXiuaSGmRK+ma2u3EliYiJ1qofR+XEjUPKRw4doVOdRQir706dHV+Li4tyyf/36dWo9GkFESDVCqlXivaFD3LLnFpI11s5nCieampzsx6M/om79hkTt2kfd+g35ePRHTtvu1qMXv8+ef9uxwYMG8uZbg1m/aRtvvfMugwe94bC9Inlz0K9BOZoOX0a9oUvw9BDahRen3aiVNHpvKY3eW8qWf88wf5shDrZm7ykaDDOOvzR5K2N6OqaTZLXErtnPJTXMlvDNTHUlJSZ+Po4KAQHJ++++/SbP/u8ltu3cxyN58zH1+2/dsu/t7c3CJcvZtO0vNm6JYvGihWzcsMEtm+6gneg9IjU52XlzZtOte08AunXvydzZzku/1qpdh3z5bpcDFhEuXbwIwMULFyhatKhTNj09hBzZPPH0EHJm9+LEhevJ53Ln8KKWvw8Loo4BcPVGYvK5XN6eOBoj22qJXSueS0qYLeGb2eqKPTHR0SxeOJ+evfsAhmTI6lUraPtYewC6du/B/LnuyRuLCLlz5wYMTaeE+Pj76qDEQxzeMiqZrk/UXk729KmTFLFV2sJFinD61ElT8hgx+hMea9Wct954nZvqJktXrHU47Ynz1/li8T9sHdGS6/GJrNx9klW7b5WreVAx1u49xeXrCbcdG/R4ZQrmyUH3zxzPK4m0JHbbd+jktL3UcOe5pEZKEr6bNm102y5k/LpyJ4NeH8DQ90dw+fIlAM6eOcMjj+RN/lEs5uvHsWPH3C5zYmIiNSJCOXjwAE8/+zwRkebLSDtKRm5hOoqV8USLi8gKEdktIrtE5EV3baYmJ2vLz7Qv5JtJExkxagx7Dx5hxMgxPP/MUw6nfSRXNpoFFSNi0HyqvT6XXN6etI+8pR31WEQJft989LY0C6KOUfudxTw5YT0D2zqnd3MvJXbdeS73msxQV+xZOH8uBQv5EBRivey1p6cnG7dGceBwNFs2b2LXzvujau7Mq3xGdrZWvs4nAK8opSoC1YHnRaSiq8ZSkpMt5FOYE8ePA3Di+HEKFvIxodjww7QptLHl8Vj7jmzd4vhgQZ1AH/6LvcKZy3EkJCrmb4shvKwxWJQ/d3aCSuVj6Y7jKabdsD+WkgUfIn/u7A7llZ7E7jffTzO18rnzXFLDCgnfzFJX7Nm4YT0L582hakBZ+vbsxppVK3jjtZe5cOE8CQnGW8uxmGiKFStmSrkB8ubNS9169Vm8eKFpNp1FO9E0UEodV0pts32+BOwBXPrrSE1OtkWr1kyfZsjaT582hZatzZF+LVK0GGtXrwJg1YrllC1X3uG00WevEVomPzmzG+vvagf4sP+40WfWKsSPpTuOcyPh1pLFUoUeSv5cpUResnt5cvZy+iOw90pi1x53nktqmC3hm5nqij1Dhn3ArgNH2LH3IN9MmU7tuvX56rup1K5Tj1m//wrAj9Om0ryle+U+ffp08oyNa9eusWzpEvz9A9JJZR1ZwYnekz5Rm+pnMHBXZ9dtksnFU5ZMTk1OdsCrA+nVrQtTv/+W4iVKMnn6DKfL9mSPJ1izZhVnYmPxL1uCQW8P4bMJXzLw1ZdJSEggR44cjPt8osP2th86y9ytMSx+uyGJiYq/j55n6ppDALQLL85nC/fedn2rED86PlqC+ETF9bhEnv7KsZFSqyV2zX4uqWG2hG9mqiuO8O77H9K35xMMH/oOVasF0cM26OQqJ44f56k+vUhMTOSmukn7Dp1o0bKVSaV1gYzrGx3GUslkABHJDawChiulfkvr2pDQMLV6vTlzD+8l5f6X5m25jQ5Acjc6AEnK5MhmTQQSKySTvQuXV77dxjp8/aFPWmZIyWRLpziJSDbgV2B6eg5Uo9E8YJg82V5EXrYNYu8UkR9FJIeIlBaRjSJyQERmikh227Xetv0DtvOlXL0NK0fnBfgG2KOU+tiqfDQaTeZEABHHtzRtifgC/weEKaUqA55AF+Aj4BOlVDngHJCkOtwXOGc7/ontOpewsiVaE+gBNBCRKNvWwsL8NBpNpkLw8HB8cwAvIKeIeAG5gONAAyAp4O9koJ3tc1vbPrbzDcXF0SsrJZPXkiW6jTUajVWYNequlIoRkdHAf8A1YDGwFTivlEpa2RLNrRlCvsBRW9oEEbkAFABinc07Uyz71Gg0WRAnXuVtvragiGyx2/onmxLJh9G6LA0UAx4Cmt2L28h0yz41Gk3WQMDR1/QkYtMYnW8EHFJKnQYQkd8wuhTzioiXrTXqB8TYro8BigPRttf/R4Azzt+FbolqNJr7iFkDSxiv8dVFJJetb7MhsBtYAXSwXdMLSIrgMtu2j+38cuXifE/dEtVoNPcNE/tEN4rIL8A2jCXn24FJwDxghoi8bzv2jS3JN8BUETkAnMUYyXcJ7UQ1Gs39wbEWpsMopYYAd0aZ/heISOHa60BHM/LVTlSj0dwXjHmimX8CT4Zzoh4WPVQnO7CdYv84a5dllrFQ1/6/L82LN3oncQnWLc3M7pV5l5R6WVgXrVrGbY3VjB1YxFEynBPVaDQPDlnAh2onqtFo7hNi7RvivUI7UY1Gc1/IKn2imWKe6DP9+1DSrzBhwVWSj509e5ZWzZtQtWIFWjV3TQL3To4ePUrTRvUJrlqRkGqVGD/O8TBdKRF99CgtmzYkPLgyESG35Hv/3vEXDevWpHpYNTq1b8NFm9CZIzzduAKrhzVl1bCmTOxfHW8vD2oH+rD0ncYsH9KYOW/Up7SPIUTmmz8Xv71Wj2VDGrPy3SY0rFLE6Xsw+5mAIdtbv1Z1akYEExlShQ/ee/e2868PeJFiBR9OObETmF12KyWTU7L9946/aFC3JpGh1ej4uHP1JC3Onz/PE507ElQ5kOAqFdm44U9T7LqCifNE7xuZwol279GbP+YsuO3YmFEjqNegATt2/0O9Bg1clga2x8vLixEjx7B9x25Wrd3AlxM/d1u+d/iIUWzevpNlq9bz1ZeGfO8Lz/Zn6PsfsGHLX7Ru046xn4x2yF6RvDnp17AcTd5bSt13FhlyzJElGNk9lGe/2kCDoUv4beN/vNwqEICXWwUye/NRGg5dQv8vN/BRd+f1e8x+JmDI9s5ZuJR1m7azduM2li5exOaNRjDqbVu3cP68+z+IYH7ZrZRMTsn2C8/2Z9h7H7Bxq62efOxYPUmP1wa8ROOmTYnauYeNW6PwDwg0xa4rZIXI9pnCidaqXYf8d0jVGhK4xoKDbt17uSSBeydFixYlOOSW3G5AQCDHjsWkkyp1UpTvPRbDwQP/ULNWHQDqN2jM7D8cD7Xq5elBjuxJcsyenDx/DYUiT45sRj45s3HivE2iWRn7AA/nysbJ89ecvgeznwncLdsbn2DI9iYmJvLOoIEMG+5yVLLbMLvsVkomp2T7wP5/qFnbqCcNGjZmlhP1JDUuXLjA2rWr6f2kEREue/bs5M2b1227rpIVWqKZtk/01KmTyRW2SJEinDJJAjeJI4cPExW1nfAIc+RkjxyxyfeGRxIQWIl5c2bRqk07/vjtF2Kij6ZvADhx/hoTFu1j+8iWXItPZNWuk6zcdZKXv9/CDy/V5npcIpeux9N8+DIARs7exU8D6tC3QTlyeXvRYcwq9+7BxGeSmJhI3Rrh/HvwAP2efo6wiEi+GD+O5i1bJ0sbm4nZ32cSVshIJxFQsRJz58yidZt2/O5EPUmLw4cOUbBgIZ7u14cdO/4iOCSE0R+P5aGHHko/sdmI7hNNE1tU6U0i8pct2vRQC/My9cu4fPkyXTu1Z9SYT++S23XVXo+uHRkxypDvnfDl13w16Qvq1Ajn0uVLZMvumLpnkhxz2MD5VH1ljuEYq5fgmcYVeOLTNQS9NpcZaw8zrHMQAI9HlmDmusMEvTaXJ8au4fN+ES7/opv9TDw9PVm7cRu7D/zHti2bWbd2NX/89gtPP/eC27bvxOyy22OljPSEL7/m6y+/oPaj4Vy+5Hg9SYuExASitm+j39PPsGHzNh566CFGj3S/K8wVzAzKfD+x8nX+BtBAKVUNCAKaiUh1s4z7+BTmuE0C9/jx4xQySQI3Pj6erp3a07lrN9o95v4k+vj4eLp37UCnzk8kS+tW8A9g1txFrF6/mQ6dulC6dFmHbNWpWNgmx3yDhETFvK3RRJQrSKXiedl26CwAf2w+Sng5Q6L5iVqlmWXTuN9y8Aw5snlSILe3S/dg5jOxJ2/evNSuW481q1by778HCK5UgSr+Zbh69SpBlSq4bd/KsoM1MtJJ+PsHMGveItb8uZkOnbtQpoxj9SQtfH398PXzI8LWIn/s8Q5ERW13265raN35NFEGl2272WybaQsfDAlcIzD19GmTTZHAVUrxzFN98Q8I5MWXB5hi7/ln+uHvH8gLdvK9p0+dAuDmzZuMGjGcvk/1T83EbcScuUpomQK35JgDC7Pv2EXy5MxGmcJGH2PdioXZf/yScf3Zq9SuWBiA8kWEYjfnAAAgAElEQVTz4J3Nk9hLN5y+BzOfCUDsHbK9K5YtJSg4hP2Hj/H3vn/5e9+/5MqVi6hd/7iVjxVlvxMrZKSTuK2efDicPv0cqydpUaRIEfz8ivPPvn0ArFi+jMDA+zmwlPlbopb2iYqIJ0Z06XLA50qpuySTHaFXjydYs3olZ2JjKV+mOG8PfpdXXnuDHk90Zsp3hgTu1B9mul3e9evW8cP0qVSuXIXIUOOVeOj7H9CsuWuqJhvWr2OGTb63ZqQxwPHO0Pc5eOAAX305AYA2bR+je88nHbK37dBZ5m6NZuk7jUm4qdj53zmmrv6X4+eu8e1zNVAKzl+J46XvNgMwZOZffNwrjGcaV0Apxf9963wryexnAnDixHGeeepJbiYmcvPmTR5r35FmLcyX7TW77FZKJqdk+8qVK0yaaKsn7R6jRy/H6kl6jPlkHE/26k58XBylSpfhy6+/NcWu02SRyfaWSyYDiEhe4Hfgf0qpnXecu6U7X6JE6N79hy0pg5VfVryFa8QByj6v187fSWZeO28lnhbV85rVw9lmsmRynuIBKuilrx2+fu2rtR88yeQklFLnMYKj3hWuXyk1SSkVppQKK1iw0L0ojkajySDoPtE0EJFCthYoIpITaAzstSo/jUaT+dB9omlTFJhs6xf1AH5SSs21MD+NRpPJyMgtTEexUjJ5BxBslX2NRpPJyeAtTEfJtCuWNBpN5kYetKDMIlIDKGWfRik1xYIyaTSaB4Qs4EMdc6IiMhUoC0QBibbDCtBOVKPRuIxVckD3EkdbomFARVd1mTUajSYlsoAPddiJ7gSKAMctLItGo3mAELFuccC9xFEnWhDYLSKbMAKLAKCUcn/BukajeWB5kAaW3rWyEEkImXMtrZentWW2cmlm8f7uxxxIjaOTOltm28qlmV6e1i7ks7JX7FpcYvoXucBNi8qcBXyoY05UKbVKRAoD4bZDm5RSp6wrlkajyeoIxjSnzE6qP7kiUsLucydgE9AR6ARsFJEO1hdPo9FkZTzE8S2jklZLNFJEOiqlxgBvAeFJrU8RKQQsBawLL6TRaLI2GTywiKOk2hJVSv0MnEi67o7X9zNppdVoNBpHyAoBSNJ0hEqp6baPC0VkkYj0FpHewDxgfuoprePpfn0oUcyH0KDKmco2QED50oQHVyUyLJia1cPTT+AEixctpGolfyoFlGOUi5o5TzeuwJr3mrF6WDO+fNrQtK8V4MOyIU1YPawZ4/tGJE9JaRZUjJVDm7Li3SYseacxkeULupanyc88+uhRWjRpSFhQZcKDqzBh/DgAzp49S5sWTQiq5E+bFk04d849Webr169T69EIIkKqEVKtEu8NHWJG8ZMxWxs+qGI5akUEUffRUBrUNqRBhrw1kMjgytSODKZHlw5csKkN3CsEY7K9o1tGxaHWpFLqNWASUNW2TVJKDbSyYKnRo1dvZs1dmOlsJ7FgyXI2btnOug2bTbOZmJjIS//3PLPmLGD7jt38PONHp/XVi+TNyVONytN42BLqvLMQTw+hffWSjO8XyVMT11PnnYUcPXOVLjVLAbBmzynqDVlE/XcX8+K3m/ikt2s/CmY/cy8vLz74aBRbonayfPV6Jk2cwN49u/l49EfUrd+QqF37qFu/IR+Pdk+W2dvbm4VLlrNp219s3BLF4kUL2bhhg0l3YY02/Kz5S1n151aWrzEEJuo1aMS6zVGs2bidsuXL88kYc6SqnSHLt0TtUUr9qpQaYNt+t7JQaVGrdh3y58+f/oUZzLaVbN60ibJly1G6TBmyZ89Ox85dmDtnltN2bte09+LqjQTiEm7y70lDKmvVrhO0CvUD4MqNhOR0uby9cHUGjNnPvEjRogQF39Ka9w8I4FhMDPPmzKZb954AdOvek7mznX8+9ogIuXMbulbx8fEkxMeb1r93r7Th6zdsjJeXMSwSFh7J8Zho0/NIjywflFlE1tr+vyQiF+22SyJy8d4UMesgIrRu0ZQakWF88/Uk0+weOxaDn1/x5H1fXz9iYmKcsnHi/DUmLNxL1KhW7PykDRevxvPH5qN4eQjVSuUDoHVYcYrlz5WcpkWIL+uHN+eHF2vz4nfmqVyaxZHDh9kRFUVYRCSnT51M1rMvXKQIp0+ddNt+YmIikaFBlCjmQ4NGjYmINEfT3l4bvnp4CM8+3Y8rV664ZVNE6NC2OQ1qRTD526/uOv/D1O9p2OQu4QlLSVqx5OiWUUmvT7SW7f88SqmH7bY8SimHBLxFxFNEtovIAx+QeemKNfy5aSt/zJnPpC8msHbN6vtdpGQeyZWNZsG+hA6cR5UBs8nl7UmH6iXp/+WfvN8lmEVvN+Ly9Xhu3rzV5Jy/LYYaby2g1/h1vPGYNf3IrnL58mW6d+3IiNEf36U1b1bLxtPTk41bozhwOJotmzexa+fO9BM5gBXa8POWrGTFus3M/G0u30z6gvVr1ySfGzPyQzw9vejY+Ql3i+404sSWUXHodV5EqotIHrv9PCLi6M/ui8AeVwqX1fD19QXAx8eH1m3bsWWzOa23YsV8iY4+mrwfExOdnJej1E3StL9k07TfFkN4uYJsOXiG1iOW0/T9pfz5z2kOnrx0V9o//zlNyUK5yZ87u9v3Ygbx8fF079KBTl2eoK1NE76QT2FOHDdCP5w4fpyChXxMyy9v3rzUrVefxYvN6du1Qhu+WDGjPhTy8aFl63Zs22r0yf8wbTKLF87jy2+n3JdXZjNf50Ukr4j8IiJ7RWSPiDwqIvlFZImI7Lf9n892rYjIOBE5ICI7RCTE1XtwtE/0C+Cy3f4V27E0ERE/oCXguKRfFuXKlStcunQp+fOypUuoWMmc1ltYeDgHDuzn8KFDxMXF8fPMGbRs5VxYg+izt2va1wn0Yf/xixTM4w0Y6pr/ax7I9ysOAlDaJ3dy2qol8uHt5cHZy3Gm3I87KKV4/ul++AcE8r8XX04+3qJVa6ZPMyI3Tp82hZat3Qv7cPr0ac7bRrOvXbvGsqVL8PcPcMtmEmZrw99Z91YsX0JgxUosW7KIzz4Zw/SZv5MrV650rJiPMTpv6mT7scBCpVQAUA2j8fYGsEwpVR5YZtsHaA6Ut239ccCfpYaja+fFPgyeUuqmiDiS9lPgdSBPahfcIZmcrsGe3buyZtVKYmNjKVvKj8HvDKV3n74OFCV9rLR96uRJunQ0WkUJCQl06tKVJk3N6YPy8vLik7Hjad2yKYmJifTq3YeKlSo5ZWPbv2eZs+Uoy4Y0ISFR8fd/55iy6iBvPlaFJtWK4eEB3684yNq9xnThVqF+dKpRioTEm1yLS+Spia5NwTH7mf+5fh0//jCNSpWrUCPCaFwMGfY+A14dSK9uXZj6/bcUL1GSydNnuJwHGK3Zp/r0IjExkZvqJu07dKJFy1Zu2bTHTG3406dO0rOrscAwISGR9p260LBxU8KqBnDjxg3atzHqYVh4JGPGTTCl/A5h4oCRiDwC1AF6Ayil4oA4EWkL1LNdNhlYCQwE2gJTbH5tg60VW1Qp5XSkOod050XkN1vmSd76OaC+UqpdGmlaAS2UUs+JSD3gVaVUmrUsNDRMrdu4xcGiZxysDrNq5WuWDkByNzoAyd00qB1J1LatplbEAmUqqRbv/eDw9dO6Bx0BYu0OTVJKTQIQkSCMaZi7MVqhWzG6EmOUUkmqwwKcU0rltY3RjFBKJQ2eLwMGKqWcdkCO1pZngBpADBANRGJrPaZBTaCNiBwGZgANRGSaswXUaDRZFyf7RGOVUmF2m/0UFy8gBPhCKRWM0eX4hn1etlan6b9gjkZxOgV0ccawUupN4E0Au5Zod2cLqNFosiZJfaImEQ1EK6U22vZ/wXCiJ5Ne00WkKJC0fD0GKG6X3s92zGkc1VjKAfQFKgE5ko4rpfq4kqlGo9GAeV1VSqkTInJURPyVUvuAhhiv9ruBXsAI2/9JqyxmAy+IyAyMN+sLrvSHguMDS1OBvUBTYBjQDSemLSmlVmL0qWo0Gg1gm2xvbn///4DpIpId+Bd4EqPL8icR6QscwQjlCUbsjxbAAeCq7VqXcNSJllNKdRSRtkqpySLyA7Am3VQajUaTBmb6UKVUFIao5p00TOFaBTxvRr6OOtF42//nRaQyRog882YrazSaB5KMvCbeURx1opNsM/0HY/Ql5LZ91mg0GpfJAj7UYSf6nVIqEVgFlLGwPBqN5gFByNhxQh3F0Xmih0Rkkog0lKzQ/tZoNPcfJ2KJZmSv42hLNABohdER+62IzAFmJM32NwuFdStRrAylZfXvynWLVqEAHJlonRxzw0+si1K18H+1LLNtH6nKCuIsXG2VFPvAbKxqMWaFNpmjke2vKqV+Uko9DgQBD2O82ms0Go3LeDixZVQcLpuI1BWRCRhrUnNwa76VRqPROI2QNSLbO7pi6TCwHfgJeE0p5V6YbY1GoyFj68k7SrpOVEQ8gW+VUsPuQXk0Gs0DQpI8SGYn3dd529Qm8wIlukBqMri///oz4cFVeDinF9u2mhNCz2yp2iSskthNTEykzqNhdG5vBBlu3rgutauHUrt6KIFli9Ot8+Mu2X2mfx9K+hUmLLhK8rHffv2ZsKDK5M7h6dTzLpEvJ9/3CkneFv9fDTqF+vJ83dL80CeMyb1D+KBdRXJ7G4MiD+fw4rPOVVnyYk0GNCzrVLmjjx6lZdOGhAdXJiLkVl0BmDhhPKHVKhIRUoXBg9wXq/38s7GEBVchLKgy48d96rY9sH2f1cPo/LjxfU764nNCKvuTL5cXZ2Jj00mdPv/s20dkWHDyVrjAI6aV3RVMDsp8X3B0dH6diIwHZmKEmAJAKbXNklLdQZIMblBwCJcuXaL2o+E0aNiIwEqVmT7zF158/lnT8kqSqv1h5s/ExcVx9epVU+wmSezmzp2b+Ph4GtStRZOmzYmsXt0tuxM/H0cF/wAuXTJ0AxcsuTXe1/OJjrRo6VoE9+49evP0sy/wVJ9eyccqVqzMDzN/5f9eeMYpW/+du0bvyUZV8RD449nqrNofS8n8uZi4+hCJCp6tU5oekSX4YvUh4hJv8tXaw5Qp+BBlCjoXcd3Ly4vhI27VlTo1jLpy6tRJ5s+dzfpN2/H29ub0qVPpG0uDXbt28t23X7N63UayZ89O21bNad6iFWXLlXPL7sTPx1EhIIBLF43vs/qjNWjWoiWtmt61ctElKvj7s3GLITWSmJhI2VJ+tGn7mCm2XSEDd3U6jKMDS0EYEZyGAWNs22irCnUnqcngBgQEUqGCv2n5WClVKxZI7MbERLN44Xx69r47mNbFixdZvWoFLVq3dcl2rdp1yJ/vdinjgMBAKvi797zDSuYj5vw1Tl68wabD50i0zSbadfwiPjYpkuvxN9kRc5G4BOenAqVYV47F8M2kibz86ut4ext5FPJxb9Xyvr17CI+IIFeuXHh5eVG7Th1m/fGbWzZjou/+PqsGBVOiZCm37KbGiuXLKFOmLCVKlrTEfnoYofDE4S2j4ugUp/opbA2sLlxK2Mvgmo0VUrX2mC2xO+j1AQwdPgIPj7u/xvlzZlG3XoO7lC7vNw0DCrF0z+m7jresXIQ/D501Na8jR2x1JTySAwf2s37dWurXfpTmjeuzdctmt2xXrFiZ9WvXcubMGa5evcqihQuIsRMLdIVBrw9g6Pspf59W8PNPM+jY2akwwabzwExxEpHCIvKNiCyw7Ve0hZZKL91hEflbRKJExO1Oy7RkcM3ACqlae8yU2F24YC4FC/kQFBya4vlffp5B+4739w/kTrw8hFplC7B83+1OtGf14iQqxeLd7r1i23P58mV6dO3IiFFGXUlISODc2bMsX72e9z74iN7du7gl0xEQGMiAV1+nTcumtGvdnKpVq+Hh6fpE94Xzbd9nSMrfp9nExcUxf+4cHm/f8Z7klxpZYcWSow7+e2ARUMy2/w/wkoNp6yulgpRSKYWocpiUZHDNxgqp2pQwQ2J345/rWThvDlUDy9K3VzfWrFpB/z49ATgTG8u2rZtp0qyFWUU2hepl8vPPqcucuxqffKxFpcLULFuAoXP3mpZPfHw83bt2oFPnJ2hjqyvFfH1p0+4xRISw8AjEw8PtgZpeT/Zl3YYtLF62irz58lG+fAWXbW3cYPs+A8rSt+ft36cVLFq4gKDgEAoXLmxZHukhTrzKZ/rXeaCgUuon4CaAUioBsG4t4h2kJoNrNmZL1dpjtsTukGEfsGv/EXbsOcg3k6dTu259Jn1rSALP+uNXmjZrSY4cOdKxcm9pHFCIJXtutTYjS+XjiQg/Bv62ixsu9H+mhFKK55/ph79/IC/Y1ZVWrduyetVKAPbv/4f4uDgKFCzoVl6nbINTR//7j9l//E6nLk+4bGvIsA/YdeAIO/Ye5Jspt3+fVvDzzPv/Kg9ZoyXq6Oj8FREpgE3kSUSqAxccSKeAxSKigC/vEJbCZuuWZHLxlCWTU5PBvXHjBq8NeJHY06fp8Fhrqlatxh9zXW/dgblStfZYLbFrz2+/zOSlAa+7ZaNXjydYs3olZ2JjKV+mOG8Pfpd8+fPzysv/R+zp0zzerhVVqwYxe55jzztHNg/CS+Vj5OL9yccGNCpHNk8PPu1kTKPadewio5YcAOCX/hE8lN0TL08PapcvyMs//83hM+nPlNiwfh0zbHWlZqRRV94Z+j49evXhuaf7EhlalezZszPx6+/cHtjr1qUDZ8+cwStbNj4eO960QUh7vpzwGeM+Hs3JkyeoFRFM46bNGffFXX9GTnHlyhWWL1vCZxMmmlRK18nIU5ccxVHJ5BDgM6AysBMoBHRQSu1IJ52vUipGRHyAJcD/lFKpRqUICQ1Tq9dvcqb8DqMDkKRMdi/ruuwbj7VO/MDKACRWTwC3MgCJt0XfZ83q4WzbusXUB+NboYp6ZsLvDl//TuPyW93tFrQCR9U+t4lIXcAfY2bCPqVUfDrJUErF2P4/JSK/AxGAdaF9NBpN5iGDT6J3FEdH5zsCOZVSu4B2wExb6zStNA+JSJ6kz0ATjFasRqPRAEZgZkf/ZVQcbfsPVkpdEpFaGKJP3wBfpJOmMLBWRP4CNgHzlFLudVhqNJosQ5Lu/IOy7DOpU64l8JVSap6IvJ9WAqXUv0A1dwqn0WiyNhnZOTqKo040RkS+BBoDH4mINxl7EYFGo8kEZOQ4oY7iqCPshDHZvqlS6jyQH3jNslJpNJoszwP1Oq+UumoLzNxcRJoB65RSiy0tmUajydpk8En0juLo6Pw7wGSgAFAQ+E5E3rayYBqNJuuTFZZ9Oton2g2oppS6DiAiI4AoIM3BJY1Go0mNpNf5zI6jTvQYhjjdddu+NxBjSYkswsoObKsldnNYJIML1klUAyx+sbZltss894tlto9MtDaykbdYNyZrVT23yqpnBm5hOkqaTlREPsNY/34B2CUiS2ynGmHM/dRoNBqXMNQ+73cp3Ce9lmhSDNDdwDIMh5oArLCyUBqN5gEgg4+6O0p6TvQHYDjQBziC8eNRAvgOGGRt0TQaTVYnIw8YOUp6nTMjgXxAaaVUqFIqBCgDPAKMsrpwGo0m65L0Op/Z44mm50RbAf2VUpeSDiilLgLPYiwBvSekJpl89uxZ2rRoQlAlf9q0aMK5c+fcyufo0aM0bVSf4KoVCalWifHjxrplLyXZ4UFvvEZwlUAiQqvRpePjyYGa3eXpfn0oUcyH0KDKpth7tn9fShcvQkRI1eRjO/6Kon6dGtSICKFOjQi2bHatW/zZ/n0o5VeYcLvnMvy9dylf2o9Hw4N5NDyYRQvmO2Xz6cblWTW0CauGNmHiU5F4e3lQO8CHJYMbseydxsweWJ9SPg8B0LNuGVa+2yT5eIWieVy6j8WLFlK1kj+VAsoxykQZGatlja0qtytkhSlO6TlRpVIIOGrTord2SNqOJMnkLVE7Wb56PZMmTmDvnt18PPoj6tZvSNSufdSt35CPR3/kdj4jRo5h+47drFq7gS8nfs6e3btdtte9R2/+mLPgtmMNGjZm8/a/2bT1L8qVL8/okR+6VeYkevTqzSw3A1Lb061HL36ffbsjGzxoIG++NZj1m7bx1jvvMnjQGy7avvu5ALzwv5f4c/N2/ty8nabNHZc2KZI3B/0alKfp+0upO2QxHh5Cu4jifNQ9hOe+3kjDYUv4beN/vNyyIgC/bfyPeu8upuGwJXy+cC9DOwc5fQ+JiYm89H/PM2vOArbv2M3PM350q67YkyRrvHHLdtZv3ELOXLlMkzW2styuYHZLVEQ8RWS7iMy17ZcWkY0ickBEZopIdttxb9v+Adv5Uq7eQ3pOdLeI3CX0IiLdAfNEcdIhNcnkeXNm0627Ubxu3Xsyd/Yst/IpWrQowSG38gkICOTYMddncqUkO9yocRO8vIyu6IjI6sTEmDNTrFbtOuTPnz/9C52wl++OsotIsh76xQsXKFq0qGm23cXTU8iR3RNPDyFXdk9OnL+OAvLkyAbAwzmzcfL8NQAuX09ITpfL28slwbrNmzZRtmw5SpcpQ/bs2enYuQtz57hX/1LCbFnje1VuRxAsUft8Edhjt/8R8IlSqhxwDkgS2OwLnLMd/8R2nUukN7D0PPCbiPQBttqOhQE5AXN+Gp3EXjL59KmTFLH9IRcuUoTTp06amk9U1HbCLZBmTmLK99/RvmMny+ybzYjRn/BYq+a89cbr3FQ3Wbpiran2v5z4OT9Mn0pIaCgffDSGfPnyOZTuxPnrfLFoH9s+asW1+ERW7TrBqt0nGTB5C9NfrMX1uEQuXU+gxQfLktM8Wb8szzSuQDYvD9qPXuV0WY8di8HPr3jyvq+vH5s2bXTaTnqYLWt8r8rtEGLuvFYR8cPoZhwODBDDeAMgSfxqMvAuRhjPtrbPAL8A40VEUnrzTo80HbxSKkYpFQkMAw7btmFKqYikqPVpISJ5ReQXEdkrIntE5FFnC2hPWpLJImLaF3L58mW6dmrPqDGfWqbbPnLEcLy8vOjStZsl9q3gm0kTGTFqDHsPHmHEyDE8/8xTptnu1/9Z/t5zgD83b6dwkaIMGviKw2kfyZWNZkG+hL8xj2qvziGXtxftq5fg6cbl6TZ2LcGvz2PGukMM63wrMuN3Kw4SOWgB7/+yg5dbmSNGaDYZRdbYSsSJDSgoIlvstv53mPsUeB2boCbGMvXzNmFNgGjA1/bZFzgKycKbF2zXO41DrWSl1HKl1Ge2bVn6KZIZCyxUSgVgxBbdk871qZKSZHIhn8KcOH4cMITgChbycdX8bfl07dSezl270e4xa6SZp075ngXz5/Ht5GmZKhTYD9OmJEsQP9a+I1u3mLfeonDhwnh6euLh4cGTfZ5iy+bNDqetE1iY/2KvcOZyHAmJinnbYogoV5BKfnnZdugsALM2HyWs7N3qnr9vPkrzIN+7jqdHsWK+REcfTd6PiYnG19d5O2lhhazxvSi3owjgKeLwBsQqpcLstmTFPhFpBZxSSm1NLT+rsGz9mYg8AtTBiIKPUirOFkbPaVKTTG7RqjXTpxmystOnTaFl6zZulVkpxTNP9cU/IJAXXx7glq3UWLxoIZ+OGcVPv84iV65cluRhFUWKFmPtauPVd9WK5ZQtV94020k/hgBzZv1OxUqOzzKIOXuVkDL5yWlbHls70Id/jl0kT85slCmcG4C6FQuz/7jRn1vaJ3dy2sZVi/LvqUt3G02HsPBwDhzYz+FDh4iLi+PnmTNo2cq9+ncnVsga34tyO4OJA0s1gTa2aHMzMF7jxwJ5RSSp29KPW8vVY4DiRhnEC2Pa5hlX7sHRtfOuUBo4jRHxqRpGn+qLSqkr9he5I5k84NWB9OrWhanff0vxEiWZPH2GWwVev24dP0yfSuXKVYgMNUZsh77/Ac2cGCm2JyXZ4dEjR3Aj7gatWzQBICIiknGfuy9d27N7V9asWklsbCxlS/kx+J2h9O7TN/2EqfBkjydYs2YVZ2Jj8S9bgkFvD+GzCV8y8NWXSUhIIEeOHC6Xu7fdc6lQpjhvDX6XNatXseOvKESEkiVLOWV726GzzN0azZLBjUi8qfj7v/NMXf0vx85d5dtna3BTKc5fjePl74wFeH0blKN2oA8JiYoLV+P4v28db/Um4eXlxSdjx9O6ZVMSExPp1bsPFStVctpOalgla2x1uZ3DvC44pdSbwJsAIlIPeFUp1U1EfgY6YDjWXkDSKNps2/6ftvPLXekPBQclk10yLBIGbABqKqU2ishY4KJSanBqaayUTPbytC7og9UBSDwsXBtnZQASK7sqMnMAEqv+5sC6Z14zMoytJksml61YTX0w3fH5wF1C/BySTLZzoq1EpAyGA80PbAe6K6VuiEgOYCoQDJwFutgkjZzGypZoNBCtlEoa+vsFcG1ioUajyZJY4fSVUiuBlbbP/2JItd95zXXAlF9Ly5pnSqkTwFER8bcdaogRyESj0WgAp0fnMyRWtkQB/gdMt60S+Bd40uL8NBpNZsHkeaL3C0udqFIqCmNyvkaj0dxG0oqlzI7VLVGNRqNJFd0S1Wg0Gjd4EIIyazQajSUYr/OZ34tqJ6rRaO4bWeBtXjtRjUZzvxBEt0TNx9OiThIrV+ZYuAAFsHbFklXPG6wdNLByVVGh7pMtsw1welovy2zHJ1hTz62q4rolqtFoNC6i+0Q1Go3GHTK4AJ2jaCeq0WjuG9qJajQajRtkhYGlTLnqKqB8acKDqxIZFkzN6uFu2bJSjjn66FFaNm1IeHBlIkJu2e7dvQs1I0OoGRlCZf8y1IwMcesezJZ6vpPPxn5CaLXKhAVVoVf3J7h+/bppts2Wer4TM+SBn20eyMZRbdg0qi3PNTekRN7uFMSfH7Vm3YjW/DGoMUXy5QSgVsXCRH/blXUjjHMDH6+aluk0SUxMpHpYMI+3beWyDUi9HgJMnDCe0GoViQipwuBBA93Kx1kEY7K9o1tGJdO2RBcsWU7BgnfLPThLkhxzUMx/YqEAACAASURBVHAIly5dovaj4TRo2IhpUydTt35DXnltIGNGfcTHoz/iveHO/RF6eXkxfMQt23VqGLa/n3YrePSgga/y8COPuH0PI0aOITjEyKdGZCgNGzUmsGJFt+wCxMTEMOHzz9j21y5y5sxJ966d+fmnGfTo2dtt22BIPT/z3Av063OXqKzbJMkDz1uwBF8/P2pVD6dVqzZOPZdAv7z0blCeem/NIy7hJr+/2YiF26IZO2cX7/8UBcAzzQJ44/FqvPTNBgD+3HuSjiOXu13+8ePG4h8YmKyw6iqp1cNTp04yf+5s1m/ajre3N6dPnXK7zM6SkfXkHSVTtkTNxEo55hRt20kwK6X4/def6dDJPQkIs6We7yQhIYFr166RkJDA1WtXKVq0mGm2zZZ6tscMeWB/30fYciCWa3GJJN5UrN1zkjYRJbh0LT75moe8vUyfAhQdHc3CBfN4sk8/t22lVg+/mTSRl199HW9vbwAK+bivUeYs4sS/jEqmdKIiQusWTakRGcY3X09KP4GDWCnHfOSIzXb4LQnm9evW4FO4MOVM1CoyW+rZ19eXl15+Bf+yJSlTohiPPPwIjRo3McW21aQkDxwT49yPy56j56kR4EP+3N7kzO5J0yBffAs8BMA7nYPZ83kHOtUqw/CftieniShfiPUftebXNxoS4JfXpbK/9spLDP9wJB4e5v6J2tfDAwf2s37dWurXfpTmjeuzdYvzMinukFVe560UqvMXkSi77aKIvGSG7aUr1vDnpq38MWc+k76YwNo1q922aaUc8+XLl+nRtSMjRt1u+5efZtCho3lCZFZIPZ87d465c2az+59/OXgkhitXrvDj9Gmm2M4M7Dt2gU9m7+SPQY35/c3G7DhyjkSbHMywmdsJfP4Xflr7L/2bBgDw16GzVHzhV2oMnMOXC/fy4yv1nc5z/ry5+BTyISQ01NR7ubMeJiQkcO7sWZavXs97H3xE7+5dLJUuuRtn2qEZ14taGdl+n1IqSCkVBIQCV4HfzbCdJPHq4+ND67bt2LLZPV0mK+WY4+Pj6d61A506P5EsNwzGK/LsWb/zeIdObpXdPh8rpJ5XLFtKyVKlKFSoENmyZaNtu8fYsGG9afatxCx54CkrDlBn0FyaDV3I+Ss3OHD89j7KmWsP0TayJACXrsVz5YYhc744KoZsXh4UyOPtVH5/rl/H3Lmz8S9Xip7durByxXKe7Nnd6XLbk1I9LObrS5t2jyEihIVHIB4enImNdSsfp3BC6TMjd53eq9f5hsBBpdQRdw1duXKFS5cuJX9etnSJU/K6d2KlHLNSiuef6Ye/fyAv2NkGWLF8KRUqBODr5+dy2e3zsUrq2a9ECTZv3MjVq1dRSrFyxXICAgJNzcMqzJIHLvhwDgD8CjxEm/CS/LzuX8oWyZN8vmVYcf45dgEAn0dyJB8PLVsQD4Ezl244ld97wz/k4OFo9h04zJTpM6hXvwHfTXG99Z9aPWzVui2rV60EYP/+f4iPi6OACYO1zqDlQRynC/BjSiduk0wukbJksj2nTp6kS0fjlzQhIYFOXbrSpGkzlwtmpRzzhvXrmGGznTSN6Z2h79O0WQt+/XkmHTp1drnc9pgt9WxPREQk7R5vT42IULy8vKgWFEyffv3dtpuE2VLP9pglDzx9QD3y5/YmPvEmA77bwIWr8Xz+dE3KF3uYmzcVR2Ov8OLXxsh8u+ql6NfIn4SbN7kel8iT49zvanKX1Ophj159eO7pvkSGViV79uxM/Pq7exok2egTzcju0TEsk0xOzsDQVzoGVFJKpTk6ExIaptZtsKZzO9FCWWOru5GyeVn3wpAZ5XutRgcguZu6NSPYZrJk8v+3d97hVVTpH/+8JBiaSu+hk4RQQwsdCUVB2gLKAgYQFGXltyrsWlZYQKRYYJVVFAsKgo1mkCpNlhohhBaUIqD07tJbeH9/zCR7gUSSe2dIbjif57nPnXvuzPece2buO+ecOed9K1WN0E9nL0/z/vUr5ktTyOQ7zZ1oibYGNt7OgBoMhrsQ/7zP3sCdMKLdSKUrbzAY7m6yQnfe1QdLIpIbaAnMcjMfg8Hgn5gHS7dBVc8DBdzMw2Aw+DGZ2TqmEb9dO28wGPwbq4Xp/1bUGFGDwZAxZPJJ9GnFGFGDwZBhZAEbaoyowWDIQLKAFTVG1GAwZBCZ27FIWjFG1GAwZBhmTNQAuBu7HeBaojtL+cDdyc6Xrya6pu1mnR+Z4ryXfU8ihixyTXv9sJauaTtNZp//mVb80imzwWDIGiT56k3L6zY6wSKyXES2i0iCiDxrp+cXkcUisst+z2eni4iMF5HdIrJFRLwOdGaMqMFgyDAc9Cd6DRikquFAPeAZEQkHXgKWqmpFYKn9GSyfHhXtVz/gfW9/gzGiBoMhw3Bq2aeqHlbVjfb2WeAnoATQAUhyyzUZ6GhvdwCmqMU6IK+IFPPmN/idEd25YweRtSOSX0UK3M+749/2Wq9/v76UDS5K3Zr/C227dctmopo2JLJWdR7p1J4zXkZbfLpfH0qXLELtiKrJabNmTqd2jSrkyRHAxrgNXpc7tVDPSYx/exz35gjghJeeyt0sO1iROJvUr03XzpaTZFVlxLDB1K5eiciaVZg44d9e6aYWHnjrls00b9qQerWr82hn785p/359KFOyCHU86iSJ8f8aS56gbOmq7zIFczHr/+onv9YPbU7PhqUJLXovX/aPJObZBkzoGUHuoAAAGlQowIwB9Yh5tgEzBtQjslzaAvyldq3MnjmdOhFVuS9noM/n0yvSY0EtK1pQRDZ4vFJ0bCsiZYAIIBYooqqH7a+OAEXs7RLAfo/DDthp6cbvjGhIaCixG+KJ3RDPmtgN5MyVi/Yd/uS1Xo/oXsyeM/+GtAH9+/HqiFHExm2mXfuOvDPuLa+0H4vuzbffLbghLTy8Cl98PZNGjZt4XWb4X6jnDZu2sew/a/jwgwn8/NN2wPrTLFvyPcHBt3dynRFlB/jgvfGEhIYlf/7i88kcPHCAH+MTiN24jU5dvHNYnRQeeH38NpauWMNHE616GdC/H8NfG8W6DfY5/Vf6z2mPFOoErPpeumRxmpyKe7LvxAU6/Xstnf69li7vruXi1USWJBxlROfKjFu4kw7vrGFJwjH6NikLwOnzV+g/OZ4O76zh5enbeP3RW415SqR2rVSqXIVpX8+gYSPfz6e3pDPG0glVre3xuiVKpYjkAWYCz6nqDXdKtZznOu5A1++MqCfLly2lXLnylCpd2muNRo2bkC/fjXf03bt20tA2FFHNWxLzrXdOqBo1bkL+m7TDKlUiJDTUu8J6kFqoZ4CXXhjIiFGv++QU2c2yHzx4gO8Xzqdn7z7JaZM+/oAXXh6cHN3S2/C9qYUH/mX3zmRj0SyqJXO8OKcpXSsAL/59IK+N9q2+61UowP6TFzj0+yXKFMzF+r2nAViz+yQtK1uNp58On+W4HWpk19FzBGUPIHvA7fNM7VoJC6tESIjv59NbBGdjLIlIdiwDOk1Vk07w0aRuuv1+zE4/CAR7HF7STks3fm1Ep3/zFY90dS5aZhJh4ZWT45PPnjWDgwf23+aIjMUz1PPc72IoXrwEVatVz+hipco/XhjI8JFjbggHvHfvHmbN/IZmjSLp0vFhftm9y+d8PMMDh1WqzDz7nH7r4DmdOyeG4sWL+1zfbaoVZd6WIwDsPnqO5uHWTeTBqkUoljfHLfu3qlKEnw6d4Wpi+hpWntdKZsCpMVGx7mCfAD+p6jiPr+YASaEEegExHuk97af09YD/enT704Xb/kSft6cbbBORL0Xk1qvBS65cucL8ud/RqfMjTkkmM2Hix3w88X0a16/DubNnyX7PPY7n4RSeoZ4DAwMZ+8YYXvnn8IwuVqosXDCXgoUKUyPixnDAVy5fJkdQDpaviqXX408woP8TPuVzc3jgCRM/5qMP36dJgzqcPefMOb1w4QJvvTGawUNf9Ukne4AQVakwi7ZaRvSVmQl0qxfMjAH1yB0UyNWb5glXKJybQQ+FMHR2Qrry+aOw4BmGcw5FGwLRQJRHmPY2wBigpYjsAlrYnwHmA3uA3cBHwF+8/QmuTbYXkRLAX4FwVb0oIt9gBaz7zAn9RQsXUCOiJkWKFLn9zukkNDSMmHnWhOhdu3ayaOH82xyRMdwc6jlh21b27dtLgzoRgNVtblyvNj+sWkeRokUzuLQWsWvXsHDedyxetIDLly5x9uwZ+vXpSfESJWlnj223bd+RZ572PlhdSuGBQ0LDiJnrcU4X+H5O9+z5hX379lK/jhUc8OCBAzSqV4sVq2LTVd+NQwqy/dAZTp67AsDe4+d5YlIcYD18ahpaKHnfIvcF8e/oCF6avpX9py6mOY+UwoJnBpxa9qmqq0jd1DZPYX8FnnEib7e784FAThEJBHJhBaxzhOlfu9OVBzh+zBo2uX79Om+OHulodEunSCnUc+UqVdm7/wgJO/eQsHMPJUqUZOW6DZnGgAIMfXUUCbt+ZctPv/DJ5Gk0btqMDydNoU3b9qy0w/euXrmCChVCvNJPLTzwDed0zEj6Pun7Oa1SpSr7Dhxl+869bN+5lxIlS7JqXVy66/vh6sWYt/l/Pcn8ua1Wsgg83awcX8daQw/35gjkg961GLdwJ/G//p5m/dTCgmcGsknaX5kV14yoqh4E3gJ+Aw5jjTl8f/N+ItIvacrCiRPH06R9/vx5li1d7Mgd9fHo7jR/oCG7du4gtHwpJn/6CdO/+YoaVcKoWS2cosWLE93rca+0e0V3p1nTBuzauYOK5YKZ/OknzImZTcVywcSuW0unjm1p/7B34Z6TQj2v+GE5DerWpEHdmo62mN0se0o8P+hF5sTMokGdGrw69BXeeW+iVzpJ4YH/s2I5DSNr0jDSqpfp33xFRNUwalUPp1ix4jzWM/3ntHd0d6LsOgmx68RXcmYPoEHFAizediw57eHqRVkwqBHzBzbi2JnLzIqznnf0qF+KUgVy0j+qfPK0qCSD+0ekdq3MiZlNaPlS/Bi7li5/akfHts6dzzSTBeKDuBYy2V5eNRPoCvwOTAdmqOrU1I7x15DJbgfbuu5iWGM3y37FpfC94O7a+WwuN3tqD72lLeEYbq2db9LA+ZDJVavX1Fnfr07z/iFFc2XKkMlududbAHtV9biqXsUKVtfAxfwMBoM/kY7pTZnZ25ObRvQ3oJ6I5LKnHzTHWoplMBgMQJbozbv3dF5VY0VkBrARyzlAPHDLCgODwXAXk5mtYxpxO2TyUGCom3kYDAZ/xXi2NxgMBp/IzGOdacUYUYPBkCFk9rHOtGKMqMFgyDiygBU1RtRgMGQYbs+xvhMYI2owGDIM/zehxogaDIaMIpNPok8rmc6IurXCMTDAj12nurd6kksuhjUOyh7gmrabyz7dDPUMsGF4K9e0K/51tiu6J35Lu8OT9OH/VjTTGVGDwXB3kOTZ3t8xRtRgMGQYWcCGGiNqMBgyjqzQEvWLgcKUwveeOnWKtq1bUS08hLatW3H69Gmf89m/fz8PtmhGRLVwalavzLvj3/FZ0y3tlOrkHy/9nYiqlahbqzp/fqQTv//u/ThWjfAKNKpbg6b1axHV2IrHM+rVoTSOjKBp/Vp0bt+aw4e987HtdOjh1HjqiT6UKl6YWjWq+KwFcOnSJaIa16NhZE3q1arGqBHDAOv3VKtUgUaRtWgUWYstmzelW9uNOnmyeQWW/7MFy4Y0Z0LfOgQFWn/3FzuEs3J4S1YMbUHfZuWT968fUpDFr0Sx/J8tmDmwcbp/gzekM9pnpsQvjGhK4XvHvjmGB6Ki2LJ9Jw9ERTH2zTGpHJ12AgMDGfPGWOK3bGfFqnVM/OA9ftq+3WddN7RTqpOo5i1ZH7+VH+M2U6FiRd56Y7RPZY6Zv4QVa+NYtjIWgAHPDWJlbDwr1sbR6qE2vDX6Na90nQ49nBrRvXoTM3ehI1oAQUFBzFmwhNWxG1m5Lo6lixex/sd1AIwY9TqrYuNYFRtHteo10q3tdJ0UzZuDvs3K03r0MqJGLCVbNqFDnZJ0rV+a4vly0mTYYpoOX8K3Gw4AcF/O7IzuVoPeE9bS7NUl9Pvox3T/Bq/IAm6c/MKIphS+d953c+jxmBXEr8djvZg7JyalQ9NFsWLFiKj5v9CyYWGVOHTIqyiqrmunVCctWrYiMNAaoakbWY+DB50pexKewc0uXLjgdV/MzdDDN+eTP/+t+XiLiJAnTx7Aill09eo1x1pIbtRJYDYhR/YAArIJObMHcPT3S/RsWpZ/zfs5eRbMSTsE85/qBjM//hAHT1+8Id1tsoAN9Q8jmhLHjh2lWLFiABQtWpRjx446qv/rvn1s2hRPHRdCy7qpncSUzz6l1YPeh3sQEbp0aE1Uo7pMnvRRcvprw4ZQNbQsM77+kpcHD3OgpBZOhR52m8TERBpF1qJi6WI0a948OfTwiGFDaFA3gpdfGMjly84YIF/q5Mjvl3h/yS7Wj2rNptfbcPbSVVb8dIzSBXPTvnZJFrzcjKkDGlC2cG4AyhXOQ95c2ZkxsDELX25Gl0hnegN/hIi1Yimtr8yK2yGTn7XDJSeIyHMu5uNY6wWs0LLdHu3Mm2Pfdjy0rJvaSbwxZiSBgYH8uVsPrzXmLf6B5avX8/WsuXzy4fusWbUSgMHDRrB1x166dO3GxxMnOFJep0IP3wkCAgJYFRtHwq5fiduwnu0J2xg6fCTrNyWwfOU6Tp8+zdtj3/A5H1/r5P5c2XmwWjEiBy8k4sX55LonkE51gwkKDODy1URaj17OtFX7GBdtha4ODBCqlspH9Ltr6D5+Nc89HEa5wnl8/h23JQs0RV0zoiJSBXgSqAtUB9qKSAWn9AsXLsLhw1aExMOHD1OoUGFHdK9evUq3RzvTtVsPOv7J2dCybmon8fmUz1gwfx6TJk/16cZSvHgJAAoVLszD7TqyMe7G2FePdO3GdzHOTOz2DD0cHlI2OfTw0SNHHNF3g7x589K4yQMsXbyIosWKISIEBQXRI7oXGzf4HifM1zppHFaY/ScvcOrcFa5dV+bHH6J2+QIc/v0i8+OtB4ILNh2iUsn7ATh8+iIrth/l4pVETp2/QuyuE4Tb37lJFrChrrZEKwGxqnpBVa8BKwDHLEebtu2YNnUyANOmTubhdu191lRVnn6yL6FhlXj2+YE+690p7SS+X7SQt8e+yTczY8iVK5fXOufPn+fs2bPJ28uXLaZSeGV+2b0reZ/5c+dQMSTU5zKDc6GH3ebE8ePJMx4uXrzID8uWUDEklCP2zVxVmffdHCpVruxzXr7WycFTF6hZNj857VVjjcIKsfvwGRZuOkRDO459/ZCC7Dl6DoCFmw9Tp0KB5PHTiDL52HXkrM+/43ZkhRhLbs4T3QaMFJECwEWgDbDBG6Fe0d1Z+Z8fOHniBBXLBTN4yDAG/f0lort3ZcqnkwguVZrPv/ja5wKvWb2aL6Z9TpUqVYmsZT1hHf7aKB5q3SbTaadUJ2+9MYbLVy7Tro21rLBu3UjGv/dBurWPHztKz25dALh2LZHOj/6Z5i0fpFf3R9m9ayfZsgnBpUrz1jvveVX23h5lDykXzCtDhtHr8b5eaf0RPR/rxsoVP3DixAnKlynJkH8Op3cf7/M5cuQw/Z/sQ+L1RPT6dTp26sJDbdrSrnULTp44gapStVp1xo1P/zCH03USv+808zYeZNErUVxLvM62/f9l6qp95MgewLt96vBk8wqcv3yNv32+EYDdR87yQ8JRlg5pzvXryher97Hj0Bmv808bmXvqUlpxLWQygIj0Bf4CnAcSgMuq+txN+/QD+gEElypV6+dd+1wpi9thcN3kuovhns3a+Vtxe+28m34cXFs7P/MFrhzf7WilR9SsrctWxaZ5//y5A++6kMmo6ieqWktVmwCngZ0p7POhqtZW1doFCxZyszgGg8HgOK4u+xSRwqp6TERKYY2H1nMzP4PB4F9k5rHOtOL22vmZ9pjoVeAZVXXLn5bBYPBDssKYqNshk+/MAlyDweB3WJPtM7oUvmO8OBkMhozDGFGDwWDwHtOdNxgMBh/ICg+W/NYBicFg8H+cXPYpIg+JyA4R2S0iL7lU5FswRtRgMGQcDllREQkA3gNaA+FANxEJd6vYnhgjajAYMgwHPdvXBXar6h5VvQJ8BXRw/QeQycZE4zfGncgdlO3XNO5eEPA9hsSd13Zb32hnHW239dOjXdrpzOM3xi3KdY8UTMchOUTE0//Gh6r6ob1dAtjv8d0BwD2HvR5kKiOqqmle9ykiG9xaR+umttv6RjvraLut73bZb4eqeu81PBNhuvMGgyErcBAI9vhc0k5zHWNEDQZDVmA9UFFEyorIPcCfgTl3IuNM1Z1PJx/efpdMqe22vtHOOtpu67td9juGql4TkQHAIiAAmKSqCXcib1f9iRoMBkNWx3TnDQaDwQeMETUYDAYfMEbUkCbEyZjUdwgRye2idlF/rBOD8/iVERWRUBGpLyLZ7WVeTuu7EhRIRCqISG0RCXJBu7KINLWdXzut3UhEogFUVZ02GiLSTkSedVLTQ7sD8LqIOBNL+0btB4HZ3DilxinteiISbb/f47B2Rfs6zObWtX434jdGVEQ6ATHAa8AnwDMicp9D2iEAqpro9MUlIm2BWcCbwGdJeTmk3Rr4EngemCIijsQYtv9keYCJwMsi8jQkG1JHrhkRaQWMALY7oXeTdlPgdSBGVY85rN3K1i4GDHJYuz3WE/MWwN9wcJWQiHQEZgAvA+OAp9xsqd9N+IURFZHsQFegr6o2xzKmwcCLvhpS28htEpEvwFlDKiINsIxnL1VthhWszxHvMiLyAPAO8ISqdgSuAFWc0FbV66p6DpiMdcNqICLPJ33nq75dL58D/VR1sYjcLyKlRSSXr9o2tYCPbe3iItJSRCJF5H5fREWkBTAB6AFUBCqJSBMHyovdk3gG6K6qvYAzQA0RKSwiORzQfgropqqdgS3A48BAEbnXx6Lf9fiFEbW5D+vCBasrNRfIDnT3tptp34kHAM8BV0RkKjjeIn1dVePt7aFAfoe69UeBp1T1R7sFGgkMEJGJItLFoa73Nayb1WSgroiME5HRYuHLtXMSK+5WMfsP/i3wPlZL3YmyX/PYngH0wTrP74lIPh90A4Ce9vzD3MAOoDI4MmZ8DcgJhNkNgweAnsDbwGAfW43XgDxAUQBVnQTsw1o739YHXQOAqvrFC2iJtQKhsf05AOgOTMWe7+qlbnGsC6wg1h9uqoNlDgDu89guCcQDhey0Ag7l8wow2N7ujeXBppADuuWBl+ztQcAF4D2Hylwd2IPlKOJJrBt6H6zhifw+alfFMnBfAY/baeWAD4AHHSh7Nvv9IeAIUNWhOukCxAHrgCF2WhTwGVDdR+2n7f9KNDDS3n4K+MSJst/NL39qia4EvgeiRaSJqiaq6hdYRrC6t6KqekhVz6nqCayLKmdSi1REaopImA/aiap6xv4owO/AKVU9LiI9gNdEJKe3+h75jFTV1+ztz7Ba7U489LgIhIrIk1h/wjFAKRF5yldhVd2M1Qoao6ofqTWEMAnIB5TyUXsr1phiJFDWTtuDdSNLs5ObP9C/br8vxBrDbOtA6xxVnYE1HroS62aLqi4D7sX38dEvgQVAMyCnqj6mqhOBIk49W7hb8Ztln6p6SUSmAYr1sCMMuAwUAQ47lMdJ20C8KSI/Y/3pmjmkfQ04JyL7RWQ00AroraoXfdEVEVG7qWF/7oxVJ4d8KjDWDUZE9gNDsEJefycizYDdvmrb+tvxeLBkl70QzpzPBVjDJ8NEJMm9YgTWjcBJNmM92HtDVRN9FVPV0yKyDHhURK4AObBuBFt81P0vME1Evky6CYhITyA/4HO572oyuimc3hdwD5Zh+wqrmxPhQh7P42A3zdYUu+y/AL8BFR0ucxDQF0gAqjioGwzU8viczYX6Fqyu/HagssPaNYFRwFgnz+dNeXwDlHFQLy/wV2AF1lpwn7ryqeSRVN+u1Mnd9PLbtfP2gx9VB54W36SbD+tPMUhVfbr7p6LfG1ivDjtHsGcwtAR+UdUdTmrb+je0eJ3WBpoCR1T1ZzfycAM368TWvxdrvP/MbXdOv3ZpILuqOtKruJvxWyPqJiKSQ1UvuaTt6h/PYDDcWYwRNRgMBh/wp6fzBoPBkOkwRtRgMBh8wBhRg8Fg8AFjRA2ZFhEpIiK9MrocBsMfYYzoXYqIJIrIJhHZJiLTfXH+ISIPiMhce7u9iKTbyYqIPG1P/k76fC/WuvFl3pbLYLgTmKfzdykick5V89jb04A4VR3n8b1gXR+3nYdre5T6m6oaZxaGuw7TEjWAtVa7goiUEZEdIjIF2AYEi0grEVkrIhvtFmuS4X1IRH4WkY1ApyQhEektIu/a20VEZLaIbLZfDez0niKyxU773E4bJiJ/s7driMg6e5/ZSZ6XROQHEXldRH4UkZ0i0vhOVpLBkBLGiN7liEgg0BrYaidVBCaoamXgPDAYaKGqNYENWD4ocwAfAe2wfHem5gx6PLBCVatjLb9MEJHKtmaUnZ6SZ/spwIuqWs0u11CP7wJVtS6W+8KhKRxrMNxRjBG9e8kpIpuwDONvWM6XAX5V1XX2dj0gHFht79sLy5tQGLBXVXfZq6+mppJHFJafUNTyaPVfO226Wl6zUNVTngeI5Tg5r6qusJMmA56Oj2fZ73FAmXT/aoPBYfzGi5PBcS6qag3PBNuv8HnPJGCxqna7ab8bjrvDXLbfEzHXryETYFqihj9iHdBQRCqAFQlArBhRPwNlRKS8vV+3VI5fCvS3jw2wW5nLgEdsj/aISH7PA+zW6mmP8c5oLG9GBkOmxBhRQ6qo6nEsT/lfisgWYC0QZjtn6QfMsx8spRYM7lmgmYhsxep+h9veq0YCK0RkM1bQtJvpheXTdQtQA3jVwZ9lMDiKmeJkMBgMPmBaymfzagAAADdJREFUogaDweADxogaDAaDDxgjajAYDD5gjKjBYDD4gDGiBoPB4APGiBoMBoMPGCNqMBgMPvD/dJmW/Eqm4pEAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x119d076d8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from collections import Counter\n",
    "from sklearn.metrics import confusion_matrix\n",
    "import itertools\n",
    "\n",
    "# Predict the values from the validation dataset\n",
    "Y_pred = model.predict(x_test)\n",
    "# Convert predictions classes to one hot vectors \n",
    "Y_pred_classes = np.argmax(Y_pred, axis = 1) \n",
    "# Convert validation observations to one hot vectors\n",
    "Y_true = np.argmax(y_test, axis = 1) \n",
    "# compute the confusion matrix\n",
    "confusion_mtx = confusion_matrix(Y_true, Y_pred_classes) \n",
    "# plot the confusion matrix\n",
    "plot_confusion_matrix(confusion_mtx, classes = range(10))"
   ]
  },
  {
   "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": "3.6.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}