U2.06 - [TALLER] - Autoencoders.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "!wget --no-cache -O init.py -q https://raw.githubusercontent.com/rramosp/2020.deeplearning/master/init.py\n",
    "from init import init; init(force_download=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'2.0.0'"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import sys\n",
    "if 'google.colab' in sys.modules:\n",
    "    print (\"setting tensorflow version in colab\")\n",
    "    %tensorflow_version 2.x\n",
    "import tensorflow as tf\n",
    "tf.__version__"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import pandas as pd\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "dimension de las imagenes y las clases (1500, 784) (1500,)\n"
     ]
    }
   ],
   "source": [
    "mnist = pd.read_csv(\"local/data/mnist1.5k.csv.gz\", compression=\"gzip\", header=None).values\n",
    "X=mnist[:,1:785]/255.\n",
    "y=mnist[:,0]\n",
    "print(\"dimension de las imagenes y las clases\", X.shape, y.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x432 with 50 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "perm = np.random.permutation(list(range(X.shape[0])))[0:50]\n",
    "random_imgs   = X[perm]\n",
    "random_labels = y[perm] \n",
    "fig = plt.figure(figsize=(10,6))\n",
    "for i in range(random_imgs.shape[0]):\n",
    "    ax=fig.add_subplot(5,10,i+1)\n",
    "    plt.imshow(random_imgs[i].reshape(28,28), interpolation=\"nearest\", cmap = plt.cm.Greys_r)\n",
    "    ax.set_title(int(random_labels[i]))\n",
    "    ax.set_xticklabels([])\n",
    "    ax.set_yticklabels([])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Autoencoder with small (sparse) activations"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 97,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.model_selection import train_test_split\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=.2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 98,
   "metadata": {},
   "outputs": [],
   "source": [
    "from tensorflow.keras import Sequential, Model\n",
    "from tensorflow.keras.layers import Dense, Dropout, Flatten, Input\n",
    "from tensorflow.keras.backend import clear_session\n",
    "from tensorflow.keras.regularizers import l1\n",
    "import tensorflow as tf"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Implement the following loss function\n",
    "\n",
    "given:\n",
    "\n",
    "- $k=50$, latent dimension\n",
    "- $\\mathbf{x}^{(i)} \\in \\mathbb{R}^{784}$\n",
    "- $e(\\mathbf{x}^{(i)}) \\in \\mathbb{R}^{k}$, the encoder\n",
    "- $d(e(\\mathbf{x}^{(i)})) \\in \\mathbb{R}^{784}$, the decoder\n",
    "\n",
    "implement the following cost function\n",
    "\n",
    "$$\\text{loss}(\\mathbf{x}^{(i)}) = \\frac{1}{m}\\sum_m \\big(\\mathbf{x}^{(i)} - d(e(\\mathbf{x}^{(i)}))\\big)^2 + \\beta \\frac{1}{k}\\sum_{k,i} e(\\mathbf{x}^{(i)})$$\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 107,
   "metadata": {},
   "outputs": [],
   "source": [
    "from tensorflow.keras import backend as K\n",
    "from tensorflow.keras.losses import mse\n",
    "\n",
    "\n",
    "def get_model(input_dim, code_size, beta=.01):\n",
    "    inputs = Input(shape=input_dim)\n",
    "    encoder = Dense(code_size, activation='relu')(inputs)\n",
    "    outputs = Dense(input_dim, activation='sigmoid')(encoder)\n",
    "\n",
    "    avg_latent_activations = ...\n",
    "    loss = ...\n",
    "\n",
    "    model = Model([inputs], [outputs])    \n",
    "    model.add_loss(loss)\n",
    "\n",
    "    model.compile(optimizer='adam')\n",
    "    return model, encoder, avg_latent_activations"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 108,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "W0220 14:44:45.825977 4384662976 training_utils.py:1444] Output dense_27 missing from loss dictionary. We assume this was done on purpose. The fit and evaluate APIs will not be expecting any data to be passed to dense_27.\n"
     ]
    }
   ],
   "source": [
    "model, encoder, avg_latent_activations = get_model(input_dim=X.shape[1], code_size=50, beta=0.05)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "try with larger layer, try with more layers"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 109,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 1200 samples\n",
      "Epoch 1/100\n",
      "1200/1200 [==============================] - 0s 171us/sample - loss: 0.2083\n",
      "Epoch 2/100\n",
      "1200/1200 [==============================] - 0s 49us/sample - loss: 0.1205\n",
      "Epoch 3/100\n",
      "1200/1200 [==============================] - 0s 45us/sample - loss: 0.1021\n",
      "Epoch 4/100\n",
      "1200/1200 [==============================] - 0s 46us/sample - loss: 0.0931\n",
      "Epoch 5/100\n",
      "1200/1200 [==============================] - 0s 46us/sample - loss: 0.0850\n",
      "Epoch 6/100\n",
      "1200/1200 [==============================] - 0s 52us/sample - loss: 0.0774\n",
      "Epoch 7/100\n",
      "1200/1200 [==============================] - 0s 55us/sample - loss: 0.0709\n",
      "Epoch 8/100\n",
      "1200/1200 [==============================] - 0s 54us/sample - loss: 0.0661\n",
      "Epoch 9/100\n",
      "1200/1200 [==============================] - 0s 54us/sample - loss: 0.0631\n",
      "Epoch 10/100\n",
      "1200/1200 [==============================] - 0s 54us/sample - loss: 0.0608\n",
      "Epoch 11/100\n",
      "1200/1200 [==============================] - 0s 48us/sample - loss: 0.0588\n",
      "Epoch 12/100\n",
      "1200/1200 [==============================] - 0s 52us/sample - loss: 0.0571\n",
      "Epoch 13/100\n",
      "1200/1200 [==============================] - 0s 49us/sample - loss: 0.0557\n",
      "Epoch 14/100\n",
      "1200/1200 [==============================] - 0s 55us/sample - loss: 0.0545\n",
      "Epoch 15/100\n",
      "1200/1200 [==============================] - 0s 55us/sample - loss: 0.0535\n",
      "Epoch 16/100\n",
      "1200/1200 [==============================] - 0s 51us/sample - loss: 0.0526\n",
      "Epoch 17/100\n",
      "1200/1200 [==============================] - 0s 48us/sample - loss: 0.0517\n",
      "Epoch 18/100\n",
      "1200/1200 [==============================] - 0s 53us/sample - loss: 0.0510\n",
      "Epoch 19/100\n",
      "1200/1200 [==============================] - 0s 50us/sample - loss: 0.0503\n",
      "Epoch 20/100\n",
      "1200/1200 [==============================] - 0s 48us/sample - loss: 0.0497\n",
      "Epoch 21/100\n",
      "1200/1200 [==============================] - 0s 47us/sample - loss: 0.0492\n",
      "Epoch 22/100\n",
      "1200/1200 [==============================] - 0s 50us/sample - loss: 0.0487\n",
      "Epoch 23/100\n",
      "1200/1200 [==============================] - 0s 52us/sample - loss: 0.0482\n",
      "Epoch 24/100\n",
      "1200/1200 [==============================] - 0s 50us/sample - loss: 0.0477\n",
      "Epoch 25/100\n",
      "1200/1200 [==============================] - 0s 48us/sample - loss: 0.0473\n",
      "Epoch 26/100\n",
      "1200/1200 [==============================] - 0s 50us/sample - loss: 0.0469\n",
      "Epoch 27/100\n",
      "1200/1200 [==============================] - 0s 48us/sample - loss: 0.0465\n",
      "Epoch 28/100\n",
      "1200/1200 [==============================] - 0s 43us/sample - loss: 0.0462\n",
      "Epoch 29/100\n",
      "1200/1200 [==============================] - 0s 48us/sample - loss: 0.0458\n",
      "Epoch 30/100\n",
      "1200/1200 [==============================] - 0s 53us/sample - loss: 0.0455\n",
      "Epoch 31/100\n",
      "1200/1200 [==============================] - 0s 54us/sample - loss: 0.0452\n",
      "Epoch 32/100\n",
      "1200/1200 [==============================] - 0s 53us/sample - loss: 0.0449\n",
      "Epoch 33/100\n",
      "1200/1200 [==============================] - 0s 49us/sample - loss: 0.0446\n",
      "Epoch 34/100\n",
      "1200/1200 [==============================] - 0s 51us/sample - loss: 0.0444\n",
      "Epoch 35/100\n",
      "1200/1200 [==============================] - 0s 48us/sample - loss: 0.0441\n",
      "Epoch 36/100\n",
      "1200/1200 [==============================] - 0s 43us/sample - loss: 0.0438\n",
      "Epoch 37/100\n",
      "1200/1200 [==============================] - 0s 47us/sample - loss: 0.0436\n",
      "Epoch 38/100\n",
      "1200/1200 [==============================] - 0s 52us/sample - loss: 0.0434\n",
      "Epoch 39/100\n",
      "1200/1200 [==============================] - 0s 53us/sample - loss: 0.0431\n",
      "Epoch 40/100\n",
      "1200/1200 [==============================] - 0s 54us/sample - loss: 0.0429\n",
      "Epoch 41/100\n",
      "1200/1200 [==============================] - 0s 54us/sample - loss: 0.0427\n",
      "Epoch 42/100\n",
      "1200/1200 [==============================] - 0s 55us/sample - loss: 0.0425\n",
      "Epoch 43/100\n",
      "1200/1200 [==============================] - 0s 54us/sample - loss: 0.0423\n",
      "Epoch 44/100\n",
      "1200/1200 [==============================] - 0s 55us/sample - loss: 0.0421\n",
      "Epoch 45/100\n",
      "1200/1200 [==============================] - 0s 55us/sample - loss: 0.0419\n",
      "Epoch 46/100\n",
      "1200/1200 [==============================] - 0s 53us/sample - loss: 0.0417\n",
      "Epoch 47/100\n",
      "1200/1200 [==============================] - 0s 55us/sample - loss: 0.0415\n",
      "Epoch 48/100\n",
      "1200/1200 [==============================] - 0s 54us/sample - loss: 0.0413\n",
      "Epoch 49/100\n",
      "1200/1200 [==============================] - 0s 50us/sample - loss: 0.0411\n",
      "Epoch 50/100\n",
      "1200/1200 [==============================] - 0s 50us/sample - loss: 0.0409\n",
      "Epoch 51/100\n",
      "1200/1200 [==============================] - 0s 48us/sample - loss: 0.0408\n",
      "Epoch 52/100\n",
      "1200/1200 [==============================] - 0s 53us/sample - loss: 0.0406\n",
      "Epoch 53/100\n",
      "1200/1200 [==============================] - 0s 57us/sample - loss: 0.0404\n",
      "Epoch 54/100\n",
      "1200/1200 [==============================] - 0s 55us/sample - loss: 0.0403\n",
      "Epoch 55/100\n",
      "1200/1200 [==============================] - 0s 57us/sample - loss: 0.0401\n",
      "Epoch 56/100\n",
      "1200/1200 [==============================] - 0s 55us/sample - loss: 0.0399\n",
      "Epoch 57/100\n",
      "1200/1200 [==============================] - 0s 56us/sample - loss: 0.0398\n",
      "Epoch 58/100\n",
      "1200/1200 [==============================] - 0s 58us/sample - loss: 0.0396\n",
      "Epoch 59/100\n",
      "1200/1200 [==============================] - 0s 60us/sample - loss: 0.0395\n",
      "Epoch 60/100\n",
      "1200/1200 [==============================] - 0s 56us/sample - loss: 0.0393\n",
      "Epoch 61/100\n",
      "1200/1200 [==============================] - 0s 48us/sample - loss: 0.0392\n",
      "Epoch 62/100\n",
      "1200/1200 [==============================] - 0s 43us/sample - loss: 0.0390\n",
      "Epoch 63/100\n",
      "1200/1200 [==============================] - 0s 48us/sample - loss: 0.0389\n",
      "Epoch 64/100\n",
      "1200/1200 [==============================] - 0s 49us/sample - loss: 0.0387\n",
      "Epoch 65/100\n",
      "1200/1200 [==============================] - 0s 48us/sample - loss: 0.0386\n",
      "Epoch 66/100\n",
      "1200/1200 [==============================] - 0s 50us/sample - loss: 0.0385\n",
      "Epoch 67/100\n",
      "1200/1200 [==============================] - 0s 47us/sample - loss: 0.0383\n",
      "Epoch 68/100\n",
      "1200/1200 [==============================] - 0s 51us/sample - loss: 0.0382\n",
      "Epoch 69/100\n",
      "1200/1200 [==============================] - 0s 56us/sample - loss: 0.0380\n",
      "Epoch 70/100\n",
      "1200/1200 [==============================] - 0s 55us/sample - loss: 0.0379\n",
      "Epoch 71/100\n",
      "1200/1200 [==============================] - 0s 55us/sample - loss: 0.0378\n",
      "Epoch 72/100\n",
      "1200/1200 [==============================] - 0s 60us/sample - loss: 0.0377\n",
      "Epoch 73/100\n",
      "1200/1200 [==============================] - 0s 53us/sample - loss: 0.0375\n",
      "Epoch 74/100\n",
      "1200/1200 [==============================] - 0s 52us/sample - loss: 0.0374\n",
      "Epoch 75/100\n",
      "1200/1200 [==============================] - 0s 47us/sample - loss: 0.0373\n",
      "Epoch 76/100\n",
      "1200/1200 [==============================] - 0s 46us/sample - loss: 0.0372\n",
      "Epoch 77/100\n",
      "1200/1200 [==============================] - 0s 46us/sample - loss: 0.0370\n",
      "Epoch 78/100\n",
      "1200/1200 [==============================] - 0s 46us/sample - loss: 0.0369\n",
      "Epoch 79/100\n",
      "1200/1200 [==============================] - 0s 46us/sample - loss: 0.0368\n",
      "Epoch 80/100\n",
      "1200/1200 [==============================] - 0s 42us/sample - loss: 0.0367\n",
      "Epoch 81/100\n",
      "1200/1200 [==============================] - 0s 45us/sample - loss: 0.0366\n",
      "Epoch 82/100\n",
      "1200/1200 [==============================] - 0s 48us/sample - loss: 0.0365\n",
      "Epoch 83/100\n",
      "1200/1200 [==============================] - 0s 46us/sample - loss: 0.0363\n",
      "Epoch 84/100\n",
      "1200/1200 [==============================] - 0s 44us/sample - loss: 0.0362\n",
      "Epoch 85/100\n",
      "1200/1200 [==============================] - 0s 46us/sample - loss: 0.0361\n",
      "Epoch 86/100\n",
      "1200/1200 [==============================] - 0s 48us/sample - loss: 0.0360\n",
      "Epoch 87/100\n",
      "1200/1200 [==============================] - 0s 45us/sample - loss: 0.0359\n",
      "Epoch 88/100\n",
      "1200/1200 [==============================] - 0s 46us/sample - loss: 0.0358\n",
      "Epoch 89/100\n",
      "1200/1200 [==============================] - 0s 44us/sample - loss: 0.0357\n",
      "Epoch 90/100\n",
      "1200/1200 [==============================] - 0s 45us/sample - loss: 0.0356\n",
      "Epoch 91/100\n",
      "1200/1200 [==============================] - 0s 47us/sample - loss: 0.0355\n",
      "Epoch 92/100\n",
      "1200/1200 [==============================] - 0s 50us/sample - loss: 0.0354\n",
      "Epoch 93/100\n",
      "1200/1200 [==============================] - 0s 55us/sample - loss: 0.0353\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 94/100\n",
      "1200/1200 [==============================] - 0s 55us/sample - loss: 0.0352\n",
      "Epoch 95/100\n",
      "1200/1200 [==============================] - 0s 55us/sample - loss: 0.0351\n",
      "Epoch 96/100\n",
      "1200/1200 [==============================] - 0s 53us/sample - loss: 0.0350\n",
      "Epoch 97/100\n",
      "1200/1200 [==============================] - 0s 47us/sample - loss: 0.0349\n",
      "Epoch 98/100\n",
      "1200/1200 [==============================] - 0s 45us/sample - loss: 0.0348\n",
      "Epoch 99/100\n",
      "1200/1200 [==============================] - 0s 49us/sample - loss: 0.0348\n",
      "Epoch 100/100\n",
      "1200/1200 [==============================] - 0s 49us/sample - loss: 0.0347\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<tensorflow.python.keras.callbacks.History at 0x14999f198>"
      ]
     },
     "execution_count": 109,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.fit(X_train, epochs=100, batch_size=32)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 118,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[array([4.4608349e-01, 0.0000000e+00, 0.0000000e+00, 0.0000000e+00,\n",
       "        0.0000000e+00, 2.8072491e-01, 3.8508105e-01, 4.1526198e-01,\n",
       "        1.7899944e-03, 0.0000000e+00, 0.0000000e+00, 4.2354482e-01,\n",
       "        0.0000000e+00, 0.0000000e+00, 3.9055079e-01, 0.0000000e+00,\n",
       "        3.2591656e-01, 6.4026034e-01, 0.0000000e+00, 4.1022152e-01,\n",
       "        5.5418462e-01, 6.2894529e-01, 0.0000000e+00, 0.0000000e+00,\n",
       "        0.0000000e+00, 0.0000000e+00, 0.0000000e+00, 2.8380924e-01,\n",
       "        0.0000000e+00, 0.0000000e+00, 3.5680971e-01, 0.0000000e+00,\n",
       "        0.0000000e+00, 5.9563463e-04, 0.0000000e+00, 4.6177214e-01,\n",
       "        0.0000000e+00, 5.4042482e-01, 0.0000000e+00, 0.0000000e+00,\n",
       "        5.9360516e-01, 0.0000000e+00, 4.7642380e-01, 0.0000000e+00,\n",
       "        5.6514001e-01, 0.0000000e+00, 3.5979575e-01, 0.0000000e+00,\n",
       "        0.0000000e+00, 0.0000000e+00], dtype=float32)]"
      ]
     },
     "execution_count": 118,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "fl = K.function([model.input], [avg_latent_activations])\n",
    "fl(X_train)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "###  test the reconstruction"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 119,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Ellipsis"
      ]
     },
     "execution_count": 119,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "..."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### show the weights"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 120,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Ellipsis"
      ]
     },
     "execution_count": 120,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "..."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### show activations in latent space\n",
    "\n",
    "you should get something similar to this, indicating a much more sparse representation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 122,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA3gAAAECCAYAAACsSHXAAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4zLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvnQurowAAIABJREFUeJzt3Xe8VNW5//HvQ5NqAykiihJQCARQRISgBEUxGmPsJUZN1CRejSUmUa+J9ZfYjSXRi1eNXew16kWF2CJIsSAYmhQVRMECCtLW748zJOd59jgzh3OOnDN83q8XL/0OM2uv2bPbYmbtx1JKAgAAAADUfw3WdwcAAAAAADWDAR4AAAAAlAkGeAAAAABQJhjgAQAAAECZYIAHAAAAAGWCAR4AAAAAlAkGeAAAAABQJhjgAQAAAECZYIAHAAAAAGWCAR4AAAAAlIlG67sDpTCztL77AAAAAADr0ccppS2KPYlv8AAAAACg7ptTypMY4AEAAABAmWCABwAAAABlggEeAAAAAJQJBngAAAAAUCYY4AEAAABAmWCABwAAAABlggEeAAAAAJQJBngAAAAAUCYY4AEAAABAmWCABwAAAABlotH67kApWrVqpQEDBvw733TTTe7vf/CDH2Re8+ijj7q87bbbunz22We7fMkllxR8/ezZszPLOPXUU13+zW9+4/Ipp5zi8tixY13u3r17ps3HH3+8YD/79u3r8qRJkzJtFNOggR/Xr1mzpsptlKtu3bq5PG3aNJebN2/ucteuXTNtvPHGGy7H7eLyyy8v2IfNN98889jixYsLvmZ9ePrpp10+8MADM8/p2bOny+PGjavVPklSmzZtXP7444+r9Pel2GabbVyeM2dOlduoC9q1a+fyhx9+uJ56Uve1aNEi89gXX3xR48tp0qSJy3vssYfLS5cudXnixIm13qd12WeaNWtWMK/LMa3Yflfs+P1NWblypcuNGzd2+be//a3Ll112Wa33qVOnTi737t0785zRo0e7HI8P8+fPd3nZsmXV7tcZZ5zh8lVXXVXw+YMHD8489uKLL1arD0cccUTmsXvuuafgazbZZBOXP/vss6LLOeaYY1x+8sknXY77+siRI11u27aty9///vczy5g+fbrL8XgRr0+ifNtF69atXf7Wt77l8hNPPOFyPI+sXr264DJrQocOHTKPxe11Q8M3eAAAAABQJhjgAQAAAECZYIAHAAAAAGXCUkrruw9FmVnd7yQAAAAA1J4JKaV+xZ7EN3gAAAAAUCYY4AEAAABAmWCABwAAAABlol7UwTMzbbTRRv/Oy5cvd3/ftGnTzGtifZD333/f5ViTLmrfvr3LCxYsKKmvhcT6Zq1atco8J9YGWrJkicuxzk+swzZhwoRMm7E+SFwXlWsMStm6eG+99ZbLffr0ySzjk08+cXnKlCkux/cV673kE2vOtWzZ0uU333zT5b333tvlWGNmiy22yCwj9mvy5MkF+zRkyBCXY30jKVun5w9/+EPBNnfaaSeXY81GSXrggQdcbtTI77pbb721y7NmzSq4zHxiG3PnznU5rv9YX+ePf/xjps1zzjmnyv2orlhHbMWKFVVuo18///P28ePHu3zUUUe5fNddd1V5GbGmV6wdFI9zcVvLV1tot912c3nMmDEu//CHP3Q51vvMJy431viKx9947Iw1RDt27JhZRjwmDRs2zOXXXnvN5Xj8iPu6lF1/0bHHHuvy3/72t4LPL8W+++7rclw3Dz74YNE2fvzjH7scP6Pdd9/d5c8//9zlfHXw4mcY60MNHDjQ5bjdxNqpX331VWYZ61JPsph4Doh1xP7xj3+4HI97xbaBchJrwcXaZHfffbfL8+bNy7QRt9ftt9/e5T//+c8uX3311S4feuihmTa32morl2P94HgOjtvvYYcd5vINN9yQWUbDhg1djseguF+eeeaZLu+1116ZNuM2H68vYr9jTcZYs06SXnjhBZd33HFHl+N1ZrxHxtSpU12O9emkbO3CeC0W30d8n/lqG8Yaf3E/++53v+tyvD7ZeeedC/69JL300ksuF6sRGG233XaZx2655RaXv/3tb7scr+dKqX8dr9tPOOEEl2Mdx169erkcr6cvuOCCzDLOO++8gsuM44Kvwzd4AAAAAFAmGOABAAAAQJlggAcAAAAAZYI6eAAAAABQ91EHDwAAAAA2JAzwAAAAAKBMMMADAAAAgDJRL+rgRZtttpnL+epfxHpwPXv2dDnWPYm1K2JdoFdeeSWzjC233NLlWINq4sSJLsc6Y88++2ymzWL199aljlWstxXrtUTNmjVzOV9dlGJibZVNN93U5cp1DaVsXaZStGjRwuVYPyfWDsn3PmIdmXPPPdflU045xeXjjjvO5Xz1t2INxmL1teK2lq+fM2bMcDnWQSlWayX+fayNI2VrTMUaMrEmVbE6QVK27tett97qcqydFz/TWG8r1vySpB49emQeq6xNmzYu33PPPS7vs88+mdfEuj9x377mmmtc/sEPfuByrJvXpUuXzDJGjRrlcjyOxVpDe+65p8v5jh9RsX0/fqbxfUrSc88953KnTp1cjvX74n5X7BgmZWsVPvXUUy7HWlijR492OR7fpWwNo0WLFrkcP+NYm2nVqlUux3qiUrbOWvwM43koLiPWZZKkN954w+VY6zRuF7E+Zfx8pOL1PasqnsukbB28WI8r7gOlbL/F9O7d2+W47uK+Hz9zKXsuivU9Y93XWGcz1jGVsrXa4mti3bC4j+Rr8+GHH3Y5rs94bIy1aWOdsVhTTZLuuOMOl6+//nqX4zkhbotxfUvZWm5x34416eL7iv3cZZddMsu49957XY41AI844giXn3nmGZdffvnlTJvlIl7jxONgXdW5c2eXY826eF2Vrz7zoEGDXI77Zaz5HK/p892jJNaKfPfdd12O57d4TROvdWM9Zyl7fIj7Wby++zp8gwcAAAAAZYIBHgAAAACUCQZ4AAAAAFAmGOABAAAAQJmg0DkAAAAA1H0UOgcAAACADUmtD/DMrKGZTTKzJ3J5WzMba2bTzWykmTUp1gYAAAAAoLhv4hu8UyVVLmRyqaSrU0pdJX0i6WffQB8AAAAAoOzVaqFzM9tK0r6S/p+kM8zMJA2VdGTuKbdJOl/SDXkb+BqDBw92+cUXX6xy34oVFT755JNdjgU/SxGLa8YCibHo87po27aty/kKQcdivFEsdvzll18WfH7jxo0zj8WCkbHNWJi7FA0bNnQ5FmBfvXp1wRyLJ1dsfl7s5/e+9z2Xf/WrX7n861//2uXdd9+9aJuXXXZZ5jmV/eEPf3A5XwHaefPmufzpp5+6/Nhjj7l80003ufy///u/LsciuJJ0+OGHu7xy5UqX+/XzP/mO/W7fvn2mzdjvWHT15z//uctjxoxxORbmzlcAOO5X9913X+Y5lZ1wwgkux3UlSdtuu63LsZhp3759XZ40aVLBZeYT99343qZNm+byzJkzi7YZP6OFCxe6PGLECJcvuugil0888cRMmxdeeKHL8di4xx57uDxy5EiXL7nkEpdPPfXUzDJeeeUVl2OB8FjM/v3333d52LBhmTZj8eLhw4e7/MILLxTsV1zfscizJO28884uxwLKUSwEHYuD15Z99tnH5Vhsupi4XcTtSMoer+PxOG7vAwYMKNqn2Gaxc1nc/idOnOhyPE9J2eN13Lfj+TCeA2Lx5Hz9iIW3jznmGJcfffRRl4cOHZpp8+abb3Z54403dnn27Nkux+NJ3KdqQtwubr/99sxzvvrqK5c7dOjgctwH4nk7iucQKVu8e+utt3Z57ty5Lu+www4ut2zZMtPm+PHjXY7XH8uWLSvYz3yuu+46l2+77TaXv/jiC5fjtcAZZ5zh8lVXXZVZRjxWxmuveE6O8u3b8XNu1MgPHVatWlWwzVLEfb1jx44uL1682OWlS5cWbTMWS4/7SFxXCxYscHnvvffOtBnf+x133OFyPB788pe/dLl///4ujxs3LrOM2K8pU6ZknlOK2v4G78+Sfitp7VG1taRPU0prt4b3JHXM90IAAAAAQNXU2gDPzPaTtDClVPmfjbJfoUh575BpZiea2XgzG5/v7wEAAAAAXm3+RHOQpP3N7PuSmkraWBXf6G1qZo1y3+JtJemDfC9OKY2QNEKiTAIAAAAAlOIbqYNnZkMknZlS2s/M7pf0YErpXjO7UdKbKaW/Fnk9AzwAqKYmTfxNi4vNdwEAAHVKna2D9ztV3HBlhirm5N1c5PkAAAAAgBJ8I9/gVRff4AFA9fENHgAA9Vqd/QYPAAAAAFALarUOHgCg7uAbOwAAyh/f4AEAAABAmWCABwAAAABlggEeAAAAAJSJejkH79BDD3X5iSeeyDxn2LBhLj/66KMuH3fccS4vXrzY5QULFric726js2fPdrl3794ujxo1yuXzzjvP5Tlz5mTabNOmjcs33XSTywMGDHD5mWeeybRRTOPGjV1euXJlldv4JmyyySYuf/bZZwWf36xZM5eXLVtW5WV2797d5WnTprm8evXqom1ss802Lg8ePNjlO++80+VBgwa5PHbs2Eybq1atKrjMo446yuW77rqraD+jYus7bt877rijy48//nimzU8++cRlM3P5T3/6k8v9+/d3uUWLFi7n2w9feOEFlzfeeGOXGzTw/4610047ufy73/0u0+bWW2/tcocOHVy+9dZbXX7sscdcjp/5K6+8klnG3Xff7XK/fv6mWJ9++qnLDRs2dDkew/KJn+lGG23k8sKFC13u0aNHpo1Zs2a5vHz5cpePPfZYlx944AGX27ZtW7C9ddG0adOCfcrnwAMPdPmhhx4q+PwddtjB5fbt22eeM2bMGJfj9hqPQdtvv73LQ4cOzbT5l7/8pWC/ionHH0k6//zzXT7hhBNcjseXuK///e9/d/mUU07JLOPyyy93+eCDD3Y5bq/x+HLddddl2tx3331djvvdGWec4fL777/v8siRI10+8cQTM8sYMWJE5rGaVmx7XZftuWXLli4vXbq0Sn0aMmRI5rG4PRez2WabuRyP9/m0atXK5SVLllRpmfm0a9fO5S+//LLgMuK6u+SSSzJtXnrppS7PmzfP5c6dO7scrw169uyZafP00093ea+99nL5oIMOcjlet8ZjVr5ryHjNEveJeM3YsWNHl/fff/9Mm1dccYXL8Zge12fcftesWeNyPL9K0rbbbutyvJ6L1/nx/JjvGql58+Yux+0iat26tctfffVV5jn77LOPy/fff7/L8Tw+efJkl+M54vPPP88sI15vxO23lP1M4hs8AAAAACgbDPAAAAAAoEwwwAMAAACAMsEADwAAAADKhOW7aUFdY2Z1v5MAAAAAUHsmpJT6FXsS3+ABAAAAQJlggAcAAAAAZYIBHgAAAACUiXpR6LxRo0auAOGiRYvc38eigFK2KGVtiEWxYz9ee+01l0sp6BmL8cbCw1OnTnV5xowZLsdCxpI0ceJEl2NByIEDB7oc+3300Ue7HAt1S9mCp7FAZJMmTVyOhRrzFVs/66yzXL7xxhtdjoWHY+HtWEg6FoqWssVI33jjDZdroih8LIg6e/bsKrcR11+nTp1cnjlzpsuxeGksdhoLlUrZotfDhg1zedSoUaV1tpJYuHz8+PEuxwKo8TOKxWNjQVUpW+A3FvCNn2lcZiywKkk/+tGPXI7FkD/44AOX476/YsUKl2OBa0l6+eWXXY6fSd++fV2O++XixYszbcYitXH9xe3inXfecblr166ZNmPx13/+858ux8KtsZBuLOYbC9RK2QK0sXDxhAkTXI774aBBgzJtdunSxeVYTDoeS+N+GpfZrVu3zDJiUeHhw4e7HD+zWJh4ypQpmTZjgd9YADzuQ3G7ufDCCzNtxnVx8sknu/zqq6+6HIuQx/eRr1DxpEmTMo9VFtdfXHebbrpp5jVxP3ruuedcvvjii11+8sknXY7nz7hupYrri8rGjh2beU4hu+22W+axHXbYweWnn37a5VjcOJ63853Hd955Z5c/+ugjl+O+HI/nsUDz0KFDM8v4zW9+k3mssp/97Gcu33zzzS7Hc7KUPcc+++yzLsd1Fd9HKR555BGXDzjgAJdj8em4D+UTz0XxvcdzQimKFYaP5/l4rRb39QULFhRdZnzNsmXLXC7luvSoo45y+a677nI5Fq+P57u4vvNtJ/G99O7d2+V4Ho/HoHhOlqS9997b5Xh8WBd77rlnwb+P5434GcdzVb5jUlw/+a65S8E3eAAAAABQJhjgAQAAAECZYIAHAAAAAGWCOngAAAAAUPdRBw8AAAAANiQM8AAAAACgTDDAAwAAAIAywQAPAAAAAMoEAzwAAAAAKBMM8AAAAACgTDDAAwAAAIAywQAPAAAAAMoEAzwAAAAAKBMM8AAAAACgTDDAAwAAAIAywQAPAAAAAMoEAzwAAAAAKBMM8AAAAACgTNTaAM/MmprZODN7w8zeNrMLco9va2ZjzWy6mY00sya11QcAAAAA2JCUNMAzs+dKeSz4StLQlFJvSX0kDTezAZIulXR1SqmrpE8k/axqXQYAAAAA5NOo0F+aWVNJzSW1MbPNJFnurzaWtGWh16aUkqSludg49ydJGirpyNzjt0k6X9IN69B3SGrUyH+EG220UeY5LVu2dPnDDz90+cQTT3R5xIgRLp933nkuX3DBBUX7tckmm7jctGnTgn3I54QTTnD5wQcfdHnx4sUub7ml3yQ/+OCDosto0aKFy1988UXR11S27777Zh776KOPXB43bpzL/fr1c3n8+PFVWqYktWnTxuWPP/64Sq/feuutM4/F9bVq1aoq9yvq2LGjy127dnV5zJgxLsdtdfvtt3d5woQJmWX06dPH5ddff71KfWzbtm3msYULF7rcoIH/t7DDDz/c5bvvvtvlbt26uTxt2rTMMuL2+umnn7q8YsUKl1u1auVykybZHz8sWbLE5cGDB7v8zDPPuFzKvt29e3eXp0+f7nKx7WTgwIEud+nSJfOcO+64o2AbcbtYunSpy0ceeaSi+JlcddVVLp9xxhkFl3nAAQe4/Mgjj2Ses88++7i8YMECl2fMmOFyr169XH7llVcybcbtMW6L6yKu89WrV7u84447uhzf65o1a1zu27dvZhmTJk2qUp969Ojhcr5zV7E2Gzdu7PLKlSur1IdyEo/pc+fOrXIbhx12mMsjR46s0uvjeUnKnpuGDRvm8qhRo1z+05/+5PJLL73k8pNPPlm0H6effrrLV199tcvxeBGPFVLNnKeLiceYeJ556KGHqtzmZptt5nLc99955x2Xly9f7nIp5/1NN93U5XjuKna8zmfAgAEuVwwh/iMeH5o1a+ZyPD9K0i677FKwzSuvvNLleHz+0Y9+lGkzHgtPOeUUl5s3b+7yVltt5fL555/vcr5z1/Dhw12Ox8FSrp+lIgM8ST+XdJoqBnMT9J8B3ueS/lKscTNrmHvdt3LPnynp05TS2i3oPUkdv+blAAAAAIAqKDjASyldI+kaMzslpXRdVRtPKa2W1MfMNpX0sKTu+Z6W77VmdqKkE/P9HQAAAAAgq9g3eJKklNJ1ZjZQUufKr0kp3V7i6z81szGSBkja1Mwa5b7F20pS3t/RpZRGSBohSWaWdxAIAAAAAPgPi79JzfskszskdZH0uqS1P95PKaVfFXjNFpJW5gZ3zST9nypusHKMpAdTSvea2Y2S3kwp/bXI8hngAQAAANiQTUgp9Sv2pJK+wZPUT1KPVMpo8D86SLotNw+vgaT7UkpPmNkUSfea2cWSJkm6uQptAgAAAAC+RqkDvMmS2kuaX2rDKaU3JWVutZVSmiWpf6ntAAAAAABKU+oAr42kKWY2ThX17SRJKaX9a6VXAAAAAIAqK3WAd35tdgIAAAAAUH2l3kXzH2a2jaSuKaVnzay5pIa12zUAAAAAQFU0KOVJZnaCpAck/U/uoY6SHqmtTgEAAAAAqq6kAZ6k/5I0SNLnkpRSmi6pbW11CgAAAABQdaXOwfsqpbTCzCRJZtZIUp2pTffwww9nHvvJT37i8vHHH+/y1Vdf7XLTpk0LLmP58uWZxwYNGuTyvvvu6/Lvf/97l1evXq1ihg4d6nKPHj1c7tq1q8unnXaay1WrZIGoXbt2Ln/44YffeB/at2+feWzjjTd2efr06QVfM3++v+HtsGHDXB41alTRfsTXzJo1y+WZM2cWbSPq29ffWHfSpElVbqOYuM/EddGkSROX1+UzHjx4sMtTp051+eOPP65ym8W2vaOPPtrll156KdPGu+++W+Xl1kXx2Pryyy+7fNxxx7l86623ZtrYZZddXB47dqzL22yzjctxfec75lfXT3/6U5eXLVuWec4999xTrWV06tQp89iXX37p8qJFi6q1jGuvvTbz2K9+5cvirr1eWKvYuWnIkCGZx2I/33rrLZdPOeUUl6+77jqXDzvsMJd33333zDJOOukkl//yl7+4HI8fF198sctxW5Sy2+Pvfvc7ly+99NIq/b0k3XDDDS6fe+65Lsd1deKJJ7o8YsSITJv10Z577pl57Nlnn61SGzvttJPL8dwmSZ988kmV2mzevLnLcZ9bX+LxIF5DPv/88y537tw508bs2bMLLiOeu+K6iOel/v2zN9IfN26cy7169XJ5ypQpLpdyPf1NOOSQQ1yO++GECRNc/uyzz6q8jHi98eKLL5b0ulK/wfuHmZ0jqZmZDZN0v6THq9JBAAAAAEDtKnWAd5akjyS9Jennkv4u6dyCrwAAAAAAfKNKvYvmGkk35f4AAAAAAOqgkgZ4ZjZIFbXwtsm9xiSllNJ2tdc1AAAAAEBVWCk35TCzdySdLmmCpH/PbEwpVW+mdonMjDuHAAAAANiQTUgp9Sv2pFLvovlZSumpanYIAAAAAFCLSh3gjTazyyU9JOmrtQ+mlCbWSq8AAAAAAFVW6gBvbUGhyl8JJklD8zwXAAAAALAelHoXze/VdkcAAACQ1bhxY5dXrly5nnoCoD4oqQ6emW1iZleZ2fjcnyvNbJPa7hwAAAAAoHSlFjq/RdISSYfm/nwu6dba6hQAAAAAoOpKnYPXJaV0UKV8gZm9XhsdAgAAAACsm1K/wVtmZt9dG3KFz5fVTpcAAACw1sqVK90fACik1G/wfinptty8O5O0WNIxtdYrAAAAAECVlXoXzdcl9TazjXP581rtFQAAAACgykq9i2ZrM7tW0hhVFD2/xsxa12rPAAAAAABVUuocvHslfSTpIEkH5/5/ZG11CgAAAABQdZZSKv4kswkppZ3CY+NTSv1qrWd+WcU7CQAAAADla0Ip469Sv8EbbWaHm1mD3J9DJT1Zvf4BAAAAAGpSqd/gLZHUQtKa3EMNJH2R+/+UUtq4drr37+XzDR4AAACADVlJ3+CVehfNVtXvDwAAAACgNpVaB09m9h1JnSu/JqX0UC30CQAAAACwDkoa4JnZLZK+I+lt/ednmkkSAzwAAAAAqCNK/QZvQEqpR632BAAAAABQLaXeRfOfZsYADwAAAADqsFK/wbtNFYO8BZK+kmSquHvmd2qtZ5U0aNBAzZs3/3deunSp+/vhw4dnXvP0009Xa5lDhgxxecyYMUVfs//++7v82GOPVXm5u+yyi8vjx493uU+fPi7PnDnT5U8//TTTZr9+/mY7LVu2dDm+t1/+8pcu33DDDV/f4ZwDDzzQ5eeff97lAw44wGUzc/m5557LtDl37lyXmzRp4vJuu+3m8rPPPuvyoYce6nLlbWitkSNHurxs2TKXO3bs6HLjxo1dXr16dabNefPmZR6rba1bt3Z50KBBLsf1+8UXX6iqfvjDH7r86KOPVrmNqKr7zCWXXJJ57KabbnI57hNHHXWUy3fddZfLjRplD4Nx23rzzTddXr58ucvxmBT347Fjx2aW0bt3b5ffeOMNl7t27eryjjvu6HL79u0zbcbPJL7moYf8r+qvueYal0899dRMmwMHDnT5lVdeyTynsoYNG7ocP7P7778/85oJEya4fMopp7i8ZMkSl1etWuXyCy+8kGlzo402cjke97baaiuXH3nkEZfjPhT3fSl77IzH2vgZvvzyyy7nO3e98847Lsf3/t3vftflW2+91eUTTjgh0+bkyZNdnjFjhsv77LOPyxMnTnT5tddec/mYY47JLGPUqFEuH3bYYQX72aVLF5eHDRuWafPLL790+fbbb3c5foZxv3vppZdc/uMf/5hZxo033ujyhRde6PI///nPgsuIxwZJOvHEE12+7bbbXI7r77TTTnO5RYsWmTbjdhE99dRTLsd1t+WWW7q88847Z9qI22v8zBYtWuTy0Ucf7XLcTqTs+fDSSy91OV4rxHPqFlts4fITTzyRWUb37t1d3nrrrV2Ox5y4bv77v/8702a8noj9ivtUPE6++uqrmTZ/8IMfuBzPd3vttZfLcd3EY8Fxxx2XWcZXX33lclx/DzzwgMvz5893ec2aNYrisXPPPfd0OR4743XTdttt5/LUqVMzy4jbY9yv4ro58sgjXX799dczbcZ9YvDgwS7H8+E555zj8k9/+tNMmxdddJHL8Zz74osvuvz73//e5XhN37dv38wy4vq89tprXY7r/+uUOsC7RdLRkt7Sf+bgAQAAAADqkFIHeHNTSlX6OsrMOkm6XVJ7VQwKR6SUrjGzzSWNVMUdOWdLOjSl9ElV2gYAAAAAZJU6wHvHzO6W9LgqfqIpqWiZhFWSfp1SmmhmrSRNMLNRko6V9FxK6RIzO0vSWZJ+t069BwAAAAD8m6WUij/J7NY8D6eUUvYHql/fxqOSrs/9GZJSmm9mHSSNSSltX+i1nTt3TpV/Ix1/+xx/Ny5lf1s7evRol+PvpePveeOcmjvvvDOzjPj79H333dflhQsXunzmmWe6nO+3zvG3yvH36yNGjHA5zqVq165dps1ic8KaNWvmcpyHFufUNGiQvTfPypUrXd58881dXrx4ccE+lCL+Djx+RusizquJ7yPae++9XR43blzmOXGeZNz2Onfu7PLs2bNdPvfcczNtxrkNkyZNcjlua1H83Xj8bX++flbVn/70p8xjcf5VXH/33XdflZbRv3//zGPxM+jZs6fLTZs2dTn+Bj7OA5Syc0TJsJjZAAAgAElEQVTj7/vnzJlTsF+xT/l+y//xxx+7XNU5u3HeqyR9/vnnLr/11lsux3l7cd5fPnFObjy2fvjhh0XbKCbOAYtzJ+IxPs4ryTcvMM7ZXbFihcsHHXSQyw8++KDLhx9+uMv55hfF+Z5x/ls8b8R5abNmzcq0GecT7rTTTi7H+YqliPvm2WefXfD5sd8//vGPiy4jzm2N87VivxcsWOBynLcjSdOnT3c5zqOOx704Py5+5rFPUnYOzbHHHutynO8S95m7774702Y8N8W5rnGeTjy+xDl8UnaOY5xLVWx7j/LN4Y2fSRTn8cVjWL5lxvnM+ebLFtKrVy+X4zFNklq1auVynKv2Tdhss81c/uST4j9Mi9v8+++/73KcHx7nHtdVcQ7pJpts4nK+65W4/cZ5kvE6NN/9D6L4mnitGz+zeK0ctz0pe0yK1wrx+rmYfMe97bf3Q6I812sTUkr94oNRSd/gpZSyMzmrwMw6S+oraaykdiml+bl255tZ2+q0DQAAAACoUFKZBDPbysweNrOFZvahmT1oZlsVf6VkZi0lPSjptJTS58WeX+l1J5rZeDMbvz7+NQYAAAAA6ptS6+DdKukxSVtK6qiKuXj5frbpmFljVQzu7qo0X+/D3E8zlftv3t+WpZRGpJT6pZT6xa/fAQAAAABZpc7Bez2l1KfYY+HvTRX18xanlE6r9PjlkhZVusnK5iml3xZafo8ePVLlOQFx7lu+eTxx3kKc5xR/Wxvn4fz85z93OdbfkbL1cWI9rlj77fHHH3f53nvvzbS53377uXz11Ve7fPLJJ2deU1m+ml5xrl8pv12uqvj76fjb5/jb/fhb52LzBfKJv+v+7LPPXI6/A8/32+hYKys+p9hv4OOcSSn7u+51mTMTxXpDsQ5brPMT51fEbSDfuojflMe6VHEOSI8ePVyOcyQl6e233y7Yj06dOrk8bdq0TBvFbLvtti6/++67VXp9vvpmcT5tnCMT52rGuRNx3l/8vCSpW7duLsf3vi5zOr4J8TPbfffdXY7zt6o7B0fKzvuN2/cHH3yQeU2x40MUj7356m0VE+eqxXpncZ/KNzczzum4+eabXY7nqjifpU2bNpk24/Yc51Wed955Lsc5vnF+6MEHH5xZRpxLEuedRXG+Yqw9K2XPu//6179cjnPuYl3TuG/nOz/GuoL5arJWFo9hcU61lD3/xbmaxeYTxfOplD2HxvlE8fgRr+1iv/PV2otzW+My4rYZ50m99957mTZjjbo47zTeRyDu67HWYaz5KmW31yiuu3i8iLVSpez1Xb75spXFGmp//etfM8+J1znxPFHVexXkq+0bP5Morr+4/ea7Xjn++ONdjtfgxa4p4zkj330hNt10U5djTee4vcZl5rsXR775moWUUr827rtR/EzjcTF+xnEeq5Q9hscaxDfffHNJc/BK/QbvYzP7sZk1zP35saTsbGVvkCpq5w01s9dzf74v6RJJw8xsuqRhuQwAAAAAqKZSyyT8VBV3v7xaUpL0iqSCN15JKb0kyb7mr/cotYMAAAAAgNKUOsC7SNIxawuS54qVX6GKgR8AAAAAoA4odQ7epJRS32KP1RYzK95JAAAAAChfNToHr4GZ/XvGf+4bvFK//QMAAAAAfANKHaRdKekVM3tAFXPwDpX0/2qtVwAAAACAKivpJ5qSZGY9JA1VxY1TnkspTanNjoVl8xNNAAAAABuykn6iWfLPLHMDum9sUFdIrMUS6wJJ0qOPPlqtZWy99dYuz507t+hrDj/8cJfHjRvncqy5NnXq1Ewbbdu2dTnW+DrwwANd/t3vfudyrDklSa+++qrLsX5IrNV0yy23uHzmmWe6nK+2yDPPPONyrGUTa/CUItY3fPrppwv2c/z48S7361d0+8/UG7riiitcvvzyywvmn/zkJ5k2r7zySpf33ntvl+fMmePy3Xff7fKuu+6aaTO+t9dee83l+LnHGlRnnXWWy4MHD84sI25rcV3EOjM9e/Z0OdYNkqRjjz3W5Y8++sjlQYMGuRzr7cTae3HdSdl9Itasi3WBLrjgApe/9a1vZdqMz4n9ip97rOkVa2LG9qRs/bJYWygeL0qpg7fDDju4HGvyxHqg3/72t12eMWNGps34mcX6csX6FY/Xsd6RlK2FFWszFauPGGsGltKvYrbZZhuX8217sW5SPJ5svPHGLnfp0sXlfHWsLrnEVw7q29dPdY81v/bcc0+Xn3/++UybsQ5V3JefffbZzGsKyVfz8kc/+lHB58TtO9aGHD58eKbNeMyPtWVjPa1YyzPW0vr+97+fWUbcTmJdqljLM9ZUi+s233LjNv/iiy+6HGuo5au7265dO5fjdjFixAiX89XerG356vetS43b6or1+2Ld17jP5Nv+43VP3Ff//ve/uxzPn7FOrJStrRmPKbFW57nnnutyfB/5PuN43Vls/Q8dOtTlyy67LPOcWKMy7ne9e/d2OdY+jOe6fF8svfzyyy7HerbxvcbzSilfVsV+fvjhhy4vWLCgaBtVtT7r2ZY6Bw8AAAAAUMcxwAMAAACAMsEADwAAAADKBAM8AAAAACgTJd9Fc31q0qRJatOmzb9zvLnDSy+9lHnNypUrXV61apXLcaJ0nAQ9ZYq/n0y86YokLVq0yOV4o4Bop512cjlOxJakmTNnFmwjatiwocutWrXKPCdOon3ooYdcbt26tctbbLGFy4sXL3b5pJNOyizj/PPPdzneTCBOXo0Txnv16pVpM94op6qTauMNPOIkXinb75iLiduRJH3wwQdVaqMUcQJ3nIBcVd26dcs8Fm9k0bFjR5cHDBjg8oMPPljl5cbtM94UIU5IjpPO892sJ05+jzdVKaZDhw6Zx+bPn+/y5ptv7nLcJ+K+PWHCBJf79++fWcabb77pcry5zujRo7+mxxXy3Vgk3uwl3hQo3nyglHUVb5wQJ+3HNuP7iDf9aNQoe2+vPn36uByPx/FmBF999ZXL8XgvZW9uFG9OEm/YEW+EE294le/4Hj/XWbNmuRy3k6VLl7qc7xhW+VwnSd/5zndcjuuzcePGLu+8886ZNuN+9t5777kcJ/7HPsQbj+TTvn17l+MxP54T4rkr9infc+L7iOeReNOEYvulJO27774uP/nkkwX7EG9Sti6OPPJIl+ONtvJdb8Rj9iuvvOLyoYce6nKxm1bEayIpux/GG1ZFcfvPd5yL23y8AVu8WVq8aVa8schTTz1VsE9Sdt+Pn1lVbyqUTzw+xBuJlKLYPlPs2Lsuip2D8ynWj9hmPH7Hm7KUIp6X48294jLy3ezo/vvvL7iM8847z+V4A8F4rZxP3O/izXf22GMPl+N1bbxOlbL7bp4bftVooXMAAAAAQB3HAA8AAAAAygQDPAAAAAAoE/ViDp6Z1f1OAgAAAEDtYQ4eAAAAAGxIGOABAAAAQJlggAcAAAAAZSJbkAgli3XC3n///RpvM9a6ijVlzj77bJdjzQ0pW2dj9913d/nggw92+aKLLnI51qjba6+9Msu44oorXI61m2Kdqlh7pV+/7M+JY53A+N6uueYal2MdlFiDKl+NwfjeL774YpeHDBni8pgxYzJtVFfLli1djp/xuoh1lWLdpVLE+maxHlF168/ls9122xVsM199qB49erh8++23u5yvdl5lsb6fJM2YMcPlI444wuW4HcT6OT/72c9cvvnmmzPL2GGHHVyO6zvWFoo1kmLdQql2tqXqqol6UfVVrPkVj1Gx7pgkjR8/3uVi55VYoyrWZZKkO++80+VY5y4qVvduv/32yzz2xBNPuBxroo0bN87luN/G2rNS9rxQrB7X5MmTXe7du7fLn3/+eWYZ8b3E2pHFxM9Yyh5zhg0b5vKoUaNc3n///V1+7LHHMm3GmmnnnHOOyyNHjnQ57mfTp0/PtBm1aNHC5Vj7MW438fyYr75cPI8Xq11YE+Ix/dVXX3V5+PDhLo8dOzbTRqw7GM8Jsd7k008/7fJBBx2UabOqtWOL7UOxhnE+cZ+J990opcZlPK/84he/cDnWi4t1CN944w2X810rxGvdWDvvnXfeKdrP9aFYjdyo2HWVlK1tGq8Fii1jLb7BAwAAAIAywQAPAAAAAMoEAzwAAAAAKBPUwQMAAACAuo86eAAAAACwIWGABwAAAABlggEeAAAAAJQJ6uABAADUI7H+XrF6nwA2LHyDBwAAAABlggEeAAAAAJQJBngAAAAAUCYY4AEAAABAmagXN1lp0KCBmjVr9u+86667ur8fPXp05jWrV68u2Gbv3r1dXrBggct9+/Z1+emnn860ceWVV7r861//uuAy18URRxzh8pw5c1weO3Zs0TaKrYuoadOmLi9fvrzoa6o64buUZfTp08fl119/vWCbbdq0cfnjjz8u+Px16ZeZuZxSKtpmfE3Pnj1dfuutt4q28dOf/rTgcp988kmXFy5c6PIhhxzi8t57751ZxvHHH+/yZZdd5vJvf/vbgn2M7zNfP2tDt27dXJ42bVq12xw4cKDLH330kcuffPKJy6eddprL5557rstHHXVUZhmzZ892+eWXX65qN2vcLrvsknms2DFm6NChLp9//vku77bbbi5vvvnmmTYOO+wwl+O6iS644AKXTzjhhMxzBg0a5PJf//pXl4cPH+5y69atXb7rrrtcfvvttzPLiNvJZptt5nKrVq1c/vTTT13u2LFjps24vtu3b+/y/PnzM6+pbIcddsg8FtdnPK7Ffi5ZsqTgMnr16pV5rFEjfzkxadKkgm0MGzbM5Xzr94MPPijYRjGdOnVyOZ7nJWnlypUux3NZu3btXI7rLt+5Lm4Hc+fOdfnss892eeLEiS6PGTMm0+ayZctcjp/zlClTXI7vPT5/1KhRmWUU2w7idVGxz1iSevToUbCfXbp0cXnmzJkF24v7g5T9TJo3b+7yumxH+++/v8v/+Mc/XP7e977ncsOGDV3OdyyN59D777/f5WuvvdblI4880uVf/vKXLsf3KUknnXSSy1dccYXLxx13nMvx+Hz55Zdn2pw3b57Lcbu4++67XT766KNdjsfrW265JbOM9957z+UVK1a4HNdFXGY+3bt3d3n69Okux2P+Rhtt5HLcb2tC3E7idiRJzz77rMvxc/7yyy9LWhbf4AEAAABAmai1AZ6Z3WJmC81scqXHNjezUWY2PfffzQq1AQAAAAAoXW1+g/c3ScPDY2dJei6l1FXSc7kMAAAAAKgBVptzZMyss6QnUko9c/lfkoaklOabWQdJY1JK25fQTu1P5AEAAACAumtCSqlfsSd903Pw2qWU5ktS7r9tv+HlAwAAAEDZqrN30TSzEyWduL77AQAAAAD1xTf9Dd6HuZ9mKvffhV/3xJTSiJRSv1K+hgQAAAAAfPMDvMckHZP7/2MkPfoNLx8AAAAAylZtlkm4R9I/JW1vZu+Z2c8kXSJpmJlNlzQslwEAAAAANaBW76JZU7iLJgAAAIANXJ28iyYAAAAAoJbU2btoAgCA+mnTTTd1+dNPP11PPQGADQ/f4AEAAABAmWCABwAAAABlggEeAAAAAJSJejkHr3///i736dMn85wvvvjC5Ucf9SX3li1b5vLq1asLtvn6668X7dfw4cNdfvrpp12+/PLLXb7qqqsybey3334u33TTTS7/4he/cPnGG28s2q/1oVEjv2ltsskmLi9atKjay2jZsqXLS5cudblJkyYux88432MdO3Z0+f3336/S3+cTt9ePP/7Y5VmzZhVtI+revbvL06ZNczm+ryFDhrg8ZsyYosuIn9lnn33mcvv27V1esGBBlduI+vXzN4Zq3ry5yy+88ELmNXH9jhs3ruAyoviZSsU/1xNOOMHluJ9uueWWLn/wwQeZNnr27OnyJ598UnCZsU8XXXRR5jmxH9/73vdcnj59ustxW9xzzz0zbT788MMun3766S6PHTv2a3pc4cEHH3T5wAMPzDyna9euLm+11VYub7311i4//vjjLsf1LUmjR492+a233nL58MMPd3nNmjUuL1myxOU33ngjs4zf//73Lh9xxBEux3lo8bySbz+M22M8jg0YMMDle+65J9NGVQ0cONDluK7iuujUqVOmjXnz5rkc59x16dLF5f3339/lG264IdPmOeec4/If/vAHl9u0aeNy3J7LSVWPpT169HB5ypQpLue7borXOaeddprLf/7zn4v2MzrooINcjueqf/3rXy5vtNFGLp933nkuP/bYY5ll7LLLLgX7efzxx7ucb1uL4jFn7ty5RV9TWePGjTOPrVy50uXBgwe7HM8BkydPdrlz584ux3OfJN13330F+7XNNtu43KCB/55nzpw5mdccfPDBLj/33HMux2uxfG1UFvd9SZo0aZLL8Xhy6KGHuhyvm1atWpVps0OHDi7H6+t7773X5TguiMuUssfbhg0bFnzNAw884PKECRNcvuCCCzLLePPNN12O11b5roPy4Rs8AAAAACgTDPAAAAAAoEwwwAMAAACAMsEADwAAAADKhKWU1ncfijIz18k42fjLL7/MvCZOZo2viZPfi00KLUWvXr1c/vzzz12OE8Lz3ZRis802czlOso2uv/56l+ONBSRpn332cTlOro43toiTmOOE2HgjASk7sTS+15EjR7ocb3CwYsWKTJtDhw51+ZRTTnF5jz32cPmAAw5w+d1333U532d8ySWXuBwn7sYbYcTJrXESr5S9QUExpdwAJd60plWrVi4Xu0FHFLd/qXgh4qZNm7p82GGHuXzbbbdlXtOtWzeXd911V5dvv/12l+NNVeLk7eXLl2eWEW8AEW8CEt1///0uV3XyfCnOOOMMl/PdUKmYOEk/3ogk3jShJsT1L+U/vhYSbza18cYbuxwnnUvZG5xUVbxRgCRtscUWLn/44YdVajPeAOHFF1+scr/ifhtzvu05ip97fE28GcGgQYMybbz88stFl1NIvIlFsRvrSNmbOcQboMQboZUiHoPizXWK3bDKzDKPxRv6xONJvMnNwoULi/YzijeYiTdziDdqiJ+plL2pVXwv8SZj8X2Uoqo3cilFXH/5zvWVxWNQvHlGsddL2Wuep556quhrionn3HVZv/F8GG84E8VzQLyuLUUpN/yqqnjjnHgjrZq4+VGzZs0K5nitPH/+/Ewbxc5d8fqilGNSHDPF7TUuM15rxeusUm7Gk8eElFK/Yk/iGzwAAAAAKBMM8AAAAACgTDDAAwAAAIAyUS/n4JVit912c/mll16Kbbqcrwh2ZXvvvXfmsWeeecblYr/Fbd26tcuHHHJIps2bb77Z5XX5zXVU7DfAG7I49yFuB3379nX5nXfecTnOpZCKF7yPhaMvvvhil7fbbrtMm3HeWJw7eNZZZ7kc50k+8sgjLsffs0vSvvvu63Lv3r1dPvPMMzOvqewnP/lJ5rE4x66q1mVOSJzTOHv27IK5rohzY2PR4Xj8iHNCpOx7O+6441yO84c233xzl+NcIUnaYYcdXI5FhF999dXMayqL8+PyzbcrNqdjXcSiwMU+9zifOfb7pJNOyrwmzneL66LYeWVdxHlo8fOJ200+xY57NSFur3H+YSlzIovtE1E8Xi9atMjldZlvW8r2W0yxbbFt27Yu55vnF4/Z+c49lXXv3t3lqVOnunzhhRdmXhMLycf33rFjR5fj9p+vyHax9RVfk6+4dDFxey4mbu9xzp6UnbcX50DGeyjEY+nixYuL9qNdu3YuxzlhM2fOLPj6bbfdNvNYvPdAMXHbyzefOd/9Iiqr6lzBeCyQsnMt474cC6GvizgHMu5Dce5rvA+ElC1UHq9Jzj//fJdvuOEGl0s57sVrwDxzi5mDBwAAAAAbEgZ4AAAAAFAmGOABAAAAQJko2zl4AACUuzifvD6c0wEA64w5eAAAAACwIWGABwAAAABlggEeAAAAAJSJbDEKAABQLzDnDgDKR03Nq+YbPAAAAAAoEwzwAAAAAKBMMMADAAAAgDLBHDwAAAAAWM9qal413+ABAAAAQJlggAcAAAAAZYIBHgAAAACUCQZ4AAAAAFAmGOABAAAAQJlggAcAAAAAZWK9DPDMbLiZ/cvMZpjZWeujDwAAAABQbr7xAZ6ZNZT0F0n7SOoh6Qgz6/FN9wMAAAAAys36KHTeX9KMlNIsSTKzeyX9UNKUr3tBo0aN1KZNm3/nAQMGuL9fsGBB5jWtW7d2+bnnnnN5u+22c3ngwIEuv/DCCy5/5zvfySzDzAr2Y+nSpS63atXK5fHjx2fa/PLLLzOPFdK1a1eXZ82alXlO7PvKlStdnjx5ssuV17Ukde/e3eUXX3wxs4wjjzzS5WeffdblQw45xOVPPvnE5bvvvjvTZrNmzVxetmyZyyeddJLLt99+u8ubb765y0OHDs0s44knnnA5rqvnn3/e5ZNPPtnl66+/PtNmtOWWW7rco4f/94xXX33V5bjdSFL//v1dHjduXMFl9u7d2+U99tjD5VdeeSXzmubNm7sc3/smm2zi8po1a1zu0KFDwT5J0rRp04o+p7Ji20A+vXr1crl9+/Yujxo1qkp9KEVcRoMG/t/O2rVrl3nNpEmTXO7cubPLsdhpXN/z5s3LtBn31djGFlts4fLbb7/tcjyeSNLYsWNdjp/zt7/9bZfjvh8NGzYs89jGG2/scpMmTVyOn3s8nsf3JUlt27Z1Oe5n22yzjcuNGvnT4cyZM13O9xkuX77c5c8++6xgm6tWrXJ5t912y7QZP5N43lixYoXL8ZwxfPjwTJvxOBc/w/he47Z29NFHuxyPDZLUr18/l+N7jcfBGTNmuDxmzJhMm/F4/NVXX7k8d+5cly+88EKXzzzzTJdbtGiRWUbcTuK1w7bbbuvyZptt5nK+Y9Idd9zhcsuWLV1evXp1wTZ+9atfZdp85JFHXH7vvfdcjp/ZDjvs4PJHH33kctznJKlp06YuT5061eV4nZTvPBLFY8r06dNdjsfK+D6ieJ6Sqn7dFD/DuF1J2XPPokWLXD7uuONcjtt73K4k6corryzYr06dOrncpUsXl2fPnl0wS9n9bPfdd3c5nnd22WUXl0ePHp1pM+5nUTyWxmPlxx9/7HK+fSZue0uWLCnYRiniNXo8x8Ztbc6cOVVeRjFxe23YsKHL8X1KUrdu3Vyu6nXTWuvjJ5odJVW+Mnkv9xgAAAAAoBrWxzd4luexlHmS2YmSTpSy/8IDAAAAAMhaHyOn9yRV/h56K0kfxCellEaklPqllPoxwAMAAACA4izO0aj1BZo1kjRN0h6S3pf0mqQjU0pvF3jNR5LmSGojqeo/xC2MNut2m/Whj7RJm7RZN9usD32kTdqkzQ2nzfrQR9qs221uk1LKTjwPvvGfaKaUVpnZyZKekdRQ0i2FBne512whSWY2PqXUr9Bzq4o263ab9aGPtEmbtFk326wPfaRN2qTNDafN+tBH2qwfbRazPubgKaX0d0l/Xx/LBgAAAIByxeQ2AAAAACgT9W2AN4I2N7g260MfaZM2abNutlkf+kibtEmbG06b9aGPtFk/2izoG7/JCgAAAACgdtS3b/AAAAAAAF+jXgzwzGy4mf3LzGaY2Vm0We02bzGzhWY2uS62V8/a7GRmo81sqpm9bWan0ma12mxqZuPM7I1cmxfUtTbrQx9D2w3NbJKZPUGbda9NM5ttZm+Z2etmNp42q93mpmb2gJm9kzs27VrX2qyF9rbPrcO1fz43s9Nos1rtnp47Fk82s3vMrClt1rk2T82193ZNfOa10WZt9LFkKaU6/UcVpRRmStpOUhNJb0jqQZvVanc3STtKmlxDn1GNtlfP2uwgacfc/7dSRY3H6n7uG3KbJqll7v8bSxoraUBdarM+9DG0fYakuyU9URPt0WbNtilptqQ2NdU/2tRtko7P/X8TSZvWtTZro4+V2m4oaYEqamXR5rq101HSu5Ka5fJ9ko6lzTrVZk9JkyU1V0VFgGclda1LbdZGH6vypz58g9df0oyU0qyU0gpJ90r6IW2uu5TSC5IWV7ed2mqvnrU5P6U0Mff/SyRNVcXBjDbXrc2UUlqai41zf6o1Ubim26wPfVzLzLaStK+k/61uW7RZO22i5pjZxqr4h7ybJSmltCKl9GldarM2+hjsIWlmSmkObVZLI0nNzKyRKi7QP6DNOtVmd0mvppS+TCmtkvQPST+qY23WRh9LVh8GeB0lzauU31M1LyI38DZRS8yss6S+qvj2hTbXva2GZva6pIWSRqWU6lyb9aGPOX+W9FtJa2qgLdqsnTaTpP8zswlmdiJtVst2kj6SdGvuZ7T/a2Yt6libtdHHyg6XdE8NtrfBtZlSel/SFZLmSpov6bOU0v/RZt1pUxXfjO1mZq3NrLmk70vqVMfarI0+lqw+DPAsz2PV/ZftDblN1AIzaynpQUmnpZQ+p811l1JanVLqI2krSf3NrGdda7M+9NHM9pO0MKU0obp9o83aaTNnUEppR0n7SPovM9uNNtdZI1X8DP+GlFJfSV9Iqu7c9Zpuszb6KEkysyaS9pd0f020t6G2aWabqeLXUttK2lJSCzP7MW3WnTZTSlMlXSpplKSnVTGFaVVdarM2+lgV9WGA9578iHcrVf+r3Q25TdQwM2usigHOXSmlh2izZuR+tjRG0vC62mYd7+MgSfub2WxV/Lx7qJndSZt1qk2llD7I/XehpIdV8dN82lw370l6r9K33w+oYjBVl9qsjT6utY+kiSmlD2uovQ21zT0lvZtS+iiltFLSQ5IG0madalMppZtTSjumlHZTxfSb6XWtzdroY6nqwwDvNUldzWzb3L/QHC7pMdpEXWBmpoq5FFNTSlfRZrXb3MLMNs39fzNVnBjeqUtt1oc+SlJK6eyU0lYppc6qOHY8n1Kq7r+a0mYNtmlmLcys1dr/l7SXKn7WQ5vrIKW0QNI8M9s+99AekqbUpTZro4+VHKGa/9njhtjmXEkDzKx57jy3h+qxjYsAAAWNSURBVCrmmNNm3WlTZtY299+tJR2oGvj8a7rN2uhjqRp9UwtaVymlVWZ2sqRnVHGHpFtSSm/T5rozs3skDZHUxszek3ReSunmutJefWpTFf+qf7Skt6xi/pQknZNS+jttrpMOkm4zs4aq+Aeo+1JK1b0dfU23WR/6iPqhnaSHK6551EjS3Smlp2mzWk6RdFfuH0VnSTquDrZZ4320ijk+wyT9vLptbehtppTGmtkDkiaq4id1kySNoM2602bOg2bWWtJKSf+VUvqkDrZZG30siaXElC4AAAAAKAf14SeaAAAAAIASMMADAAAAgDLBAA8AAAAAygQDPAAAAAAoEwzwAAAAAKBMMMADANRrZjbGzPpVs43OZlatOmzfNDMbYmaU0wAAOAzwAADYAOXqLgIAygwDPABAjTKzFmb2pJm9YWaTzeyw3ON/MLPXco+NsFyV69w3cFeb2QtmNtXMdjazh8xsupldnHtOZzN7x8xuM7M3zeyBXIHjuOy9zOyfZjbRzO43s5Z5nvMtM3s217+JZtYl/H1nM3sx93cTzWxg7vEOuT6+nnsPg82soZn9LZffMrPT8yzvb2Z2rZm9YmazzOzg3OPuGzgzu97Mjs39/2wz+2PuvYw3sx3N7Bkzm2lmv6jU/MZm9rCZTTGzG82sQaH1kGv3D2b2kqRDqvTBAgDqBQZ4AICaNlzSByml3imlnpKezj1+fUpp59xjzSTtV+k1K1JKu0m6UdKjkv5LUk9Jx5pZ69xztpc0IqX0HUmfSzqp8kLNrI2kcyXtmVLaUdJ4SWfk6d9dkv6SUuotaaCk+eHvF0oalmvjMEnX5h4/UtIzKaU+knpLel1SH0kdU0o9U0q9JN36Neukg6Tv5t7zJV/znGheSmlXSS9K+pukgyUNkHRhpef0l/RrSb0kdZF0YAnrYXlK6bsppXtL7AcAoB5ptL47AAAoO29JusLMLpX0RErpxdzj3zOz30pqLmlzSW9Lejz3d49Veu3bKaX5kmRmsyR1kvSpKgY8L+eed6ekX0m6otJyB0jqIenl3JeDTST9s3LHzKyVKgZkD0tSSml57vHKT2ss6Xoz6yNptaRuucdfk3SLmTWW9EhK6fVc/7Yzs+skPSnp/75mnTySUlojaYqZtfua50SV10nLlNISSUvMbLmZbZr7u3EppVm593CPKgaRy4ush5ElLh8AUA8xwAMA1KiU0jQz20nS9yX9ycz+T9Jlkv4qqV9KaZ6ZnS+paaWXfZX775pK/782rz1XpbiokE3SqJTSEQW6ZwX+bq3TJX2oim/pGqhiwKSU0gtmtpukfSXdYWaXp5RuN7PekvZWxbeOh0r6aZ42K7+ntX1YJf9Lmsrro/JrqrpOiq2HL77mcQBAGeAnmgCAGmVmW0r6MqV0pyq+YdtR/xm8fJybD3bwOjS9tZntmvv/IyS9FP7+VUmDzOxbuX40N7NulZ+QUvpc0ntmdkDuORvlmcu3iaT5uW/cjpbUMPfcbSQtTCndJOlmSTvmfg7ZIKX0oKTf595rqeZI6pHrwyaS9qjCa9fqb2bb5ubeHaaKdVJ0PQAAyhff4AEAalovSZeb2RpJKyX9MqX0qZndpIqfG85Wxc8dq2qqpGPM7H8kTZd0Q+W/TCl9lLtJyT1mtlHu4XMlTQvtHC3pf8zswlz/DlHFt2Jr/VXSg2Z2iKTR+s83XkMk/cbMVkpaKuknkjpKunXtzU0knV3qm8l9k3mfpDdz72dSqa+t5J+qmNPXS9ILkh5OKa0pcT0AAMqQpRR/3QEAQN1iZp1VMZ+v53ruCgAAdRo/0QQAAACAMsE3eAAAAABQJvgGDwAAAADKBAM8AAAAACgTDPAAAAAAoEwwwAMAAACAMsEADwAAAADKBAM8AAAAACgT/x9ruHwXczVLdwAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "execution_count": 122,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from IPython.display import Image\n",
    "Image(filename='local/imgs/ae_sparse_activations.png')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "..."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### show the distribution of activations in latent space\n",
    "\n",
    "should be quite concentrated at 0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "..."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### show most representative neuron in latent space for each class\n",
    "\n",
    "now, representative neurons must show somewhat distinctive shapes for each number\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "..."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "p37",
   "language": "python",
   "name": "p37"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}