{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Deep Learning with Python ３章勉強ノート（その２）\n",
    "ManningのMEAP（出版前の本のできた部分から読めるサービス）でTensorFlowとTheanoの両方に対応したラッパーツールkerasの作者Francois Chollet氏の「Deep Learning with Python」が発売されました。\n",
    "\n",
    "4章では、"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 前準備\n",
    "jupyterノートブックで使用するnumpy、pandas、seabornとmatplotlibをインポートします。\n",
    "次に、kerasで使うパッケージをインポートします。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "# 可視化用のライブラリをインポート（numpy, pandasを含む）\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import seaborn\n",
    "import matplotlib.pyplot as plt \n",
    "%matplotlib inline\n",
    "\n",
    "# keras用のパッケージをインポート\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Dense\n",
    "from keras.utils.np_utils import to_categorical"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 過学習との戦い\n",
    "３章の二値分類の例題を使ってモデルを変更すると過学習の様子がどのように変化するか試してみます。\n",
    "\n",
    "最初にもう一度データを準備します。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Downloading data from https://s3.amazonaws.com/text-datasets/imdb_full.pkl\n"
     ]
    }
   ],
   "source": [
    "# imdbデータの準備\n",
    "from keras.datasets import imdb\n",
    "(train_data, train_labels), (test_data, test_labels) = imdb.load_data(nb_words=10000)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def vectorize_sequences(sequences, dimension=10000):\n",
    "    results = np.zeros((len(sequences), dimension))\n",
    "    # create an all-zero matrix of shape (len(sequences), dimension)\n",
    "    for i, sequence in enumerate(sequences):\n",
    "        results[i, sequence] = 1.  # set specific indices of results[i] to 1s\n",
    "    return results\n",
    "# our vectorized training data\n",
    "x_train = vectorize_sequences(train_data)\n",
    "# our vectorized test data\n",
    "x_test = vectorize_sequences(test_data)\n",
    "# our vectorized labels:\n",
    "y_train = np.asarray(train_labels).astype('float32')\n",
    "y_test = np.asarray(test_labels).astype('float32')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### オリジナルのモデルを実行\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "model = Sequential()\n",
    "model.add(Dense(16, activation='relu', input_dim=10000))\n",
    "model.add(Dense(16, activation='relu'))\n",
    "model.add(Dense(1, activation='sigmoid'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "model.compile(optimizer='rmsprop', loss='binary_crossentropy', metrics=['accuracy'])\n",
    "x_val = x_train[:10000]\n",
    "partial_x_train = x_train[10000:]\n",
    "y_val = y_train[:10000]\n",
    "partial_y_train = y_train[10000:]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 15000 samples, validate on 10000 samples\n",
      "Epoch 1/20\n",
      "15000/15000 [==============================] - 218s - loss: 0.5038 - acc: 0.7880 - val_loss: 0.3868 - val_acc: 0.8673\n",
      "Epoch 2/20\n",
      "15000/15000 [==============================] - 3s - loss: 0.3013 - acc: 0.9057 - val_loss: 0.3267 - val_acc: 0.8742\n",
      "Epoch 3/20\n",
      "15000/15000 [==============================] - 2s - loss: 0.2233 - acc: 0.9256 - val_loss: 0.2770 - val_acc: 0.8939\n",
      "Epoch 4/20\n",
      "15000/15000 [==============================] - 2s - loss: 0.1754 - acc: 0.9452 - val_loss: 0.2730 - val_acc: 0.8922\n",
      "Epoch 5/20\n",
      "15000/15000 [==============================] - 3s - loss: 0.1451 - acc: 0.9542 - val_loss: 0.2803 - val_acc: 0.8891\n",
      "Epoch 6/20\n",
      "15000/15000 [==============================] - 2s - loss: 0.1166 - acc: 0.9649 - val_loss: 0.2988 - val_acc: 0.8857\n",
      "Epoch 7/20\n",
      "15000/15000 [==============================] - 2s - loss: 0.1021 - acc: 0.9691 - val_loss: 0.3045 - val_acc: 0.8845\n",
      "Epoch 8/20\n",
      "15000/15000 [==============================] - 2s - loss: 0.0826 - acc: 0.9774 - val_loss: 0.3379 - val_acc: 0.8758\n",
      "Epoch 9/20\n",
      "15000/15000 [==============================] - 2s - loss: 0.0725 - acc: 0.9797 - val_loss: 0.3572 - val_acc: 0.8803\n",
      "Epoch 10/20\n",
      "15000/15000 [==============================] - 2s - loss: 0.0579 - acc: 0.9863 - val_loss: 0.3685 - val_acc: 0.8787\n",
      "Epoch 11/20\n",
      "15000/15000 [==============================] - 2s - loss: 0.0504 - acc: 0.9879 - val_loss: 0.3946 - val_acc: 0.8766\n",
      "Epoch 12/20\n",
      "15000/15000 [==============================] - 2s - loss: 0.0400 - acc: 0.9919 - val_loss: 0.4247 - val_acc: 0.8761\n",
      "Epoch 13/20\n",
      "15000/15000 [==============================] - 2s - loss: 0.0329 - acc: 0.9940 - val_loss: 0.4476 - val_acc: 0.8745\n",
      "Epoch 14/20\n",
      "15000/15000 [==============================] - 2s - loss: 0.0274 - acc: 0.9956 - val_loss: 0.4632 - val_acc: 0.8724\n",
      "Epoch 15/20\n",
      "15000/15000 [==============================] - 2s - loss: 0.0223 - acc: 0.9966 - val_loss: 0.5818 - val_acc: 0.8608\n",
      "Epoch 16/20\n",
      "15000/15000 [==============================] - 2s - loss: 0.0158 - acc: 0.9981 - val_loss: 0.5357 - val_acc: 0.8707\n",
      "Epoch 17/20\n",
      "15000/15000 [==============================] - 2s - loss: 0.0149 - acc: 0.9983 - val_loss: 0.5677 - val_acc: 0.8694\n",
      "Epoch 18/20\n",
      "15000/15000 [==============================] - 3s - loss: 0.0125 - acc: 0.9976 - val_loss: 0.5816 - val_acc: 0.8684\n",
      "Epoch 19/20\n",
      "15000/15000 [==============================] - 3s - loss: 0.0101 - acc: 0.9982 - val_loss: 0.6077 - val_acc: 0.8665\n",
      "Epoch 20/20\n",
      "15000/15000 [==============================] - 3s - loss: 0.0055 - acc: 0.9999 - val_loss: 0.6488 - val_acc: 0.8649\n"
     ]
    }
   ],
   "source": [
    "org_history = model.fit(partial_x_train, partial_y_train, nb_epoch=20, batch_size=512, validation_data=(x_val, y_val))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 15000 samples, validate on 10000 samples\n",
      "Epoch 1/20\n",
      " 1024/15000 [=>............................] - ETA: 68s - loss: 0.6900 - acc: 0.5586 "
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/lib/sagemath/local/lib/python2.7/site-packages/keras/callbacks.py:119: UserWarning: Method on_batch_end() is slow compared to the batch update (0.148245). Check your callbacks.\n",
      "  % delta_t_median)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "15000/15000 [==============================] - 92s - loss: 0.5601 - acc: 0.7717 - val_loss: 0.4623 - val_acc: 0.8370\n",
      "Epoch 2/20\n",
      "15000/15000 [==============================] - 4s - loss: 0.3849 - acc: 0.8919 - val_loss: 0.3657 - val_acc: 0.8772\n",
      "Epoch 3/20\n",
      "15000/15000 [==============================] - 2s - loss: 0.2946 - acc: 0.9147 - val_loss: 0.3192 - val_acc: 0.8853\n",
      "Epoch 4/20\n",
      "15000/15000 [==============================] - 2s - loss: 0.2389 - acc: 0.9305 - val_loss: 0.2967 - val_acc: 0.8870\n",
      "Epoch 5/20\n",
      "15000/15000 [==============================] - 2s - loss: 0.2006 - acc: 0.9401 - val_loss: 0.2802 - val_acc: 0.8898\n",
      "Epoch 6/20\n",
      "15000/15000 [==============================] - 2s - loss: 0.1726 - acc: 0.9484 - val_loss: 0.2854 - val_acc: 0.8859\n",
      "Epoch 7/20\n",
      "15000/15000 [==============================] - 2s - loss: 0.1520 - acc: 0.9549 - val_loss: 0.2893 - val_acc: 0.8835\n",
      "Epoch 8/20\n",
      "15000/15000 [==============================] - 2s - loss: 0.1330 - acc: 0.9621 - val_loss: 0.2822 - val_acc: 0.8869\n",
      "Epoch 9/20\n",
      "15000/15000 [==============================] - 2s - loss: 0.1180 - acc: 0.9674 - val_loss: 0.3035 - val_acc: 0.8836\n",
      "Epoch 10/20\n",
      "15000/15000 [==============================] - 2s - loss: 0.1042 - acc: 0.9728 - val_loss: 0.2959 - val_acc: 0.8867\n",
      "Epoch 11/20\n",
      "15000/15000 [==============================] - 3s - loss: 0.0920 - acc: 0.9765 - val_loss: 0.3094 - val_acc: 0.8848\n",
      "Epoch 12/20\n",
      "15000/15000 [==============================] - 2s - loss: 0.0816 - acc: 0.9803 - val_loss: 0.3179 - val_acc: 0.8824\n",
      "Epoch 13/20\n",
      "15000/15000 [==============================] - 2s - loss: 0.0720 - acc: 0.9827 - val_loss: 0.3397 - val_acc: 0.8817\n",
      "Epoch 14/20\n",
      "15000/15000 [==============================] - 2s - loss: 0.0638 - acc: 0.9857 - val_loss: 0.3526 - val_acc: 0.8757\n",
      "Epoch 15/20\n",
      "15000/15000 [==============================] - 2s - loss: 0.0562 - acc: 0.9885 - val_loss: 0.3604 - val_acc: 0.8773\n",
      "Epoch 16/20\n",
      "15000/15000 [==============================] - 2s - loss: 0.0498 - acc: 0.9903 - val_loss: 0.3792 - val_acc: 0.8781\n",
      "Epoch 17/20\n",
      "15000/15000 [==============================] - 2s - loss: 0.0430 - acc: 0.9919 - val_loss: 0.3958 - val_acc: 0.8769\n",
      "Epoch 18/20\n",
      "15000/15000 [==============================] - 2s - loss: 0.0376 - acc: 0.9937 - val_loss: 0.4238 - val_acc: 0.8766\n",
      "Epoch 19/20\n",
      "15000/15000 [==============================] - 2s - loss: 0.0325 - acc: 0.9951 - val_loss: 0.4300 - val_acc: 0.8724\n",
      "Epoch 20/20\n",
      "15000/15000 [==============================] - 2s - loss: 0.0284 - acc: 0.9955 - val_loss: 0.4501 - val_acc: 0.8728\n"
     ]
    }
   ],
   "source": [
    "# 特徴ベクトルを小さく（4）にしたモデルを使って分類する\n",
    "model = Sequential()\n",
    "model.add(Dense(4, activation='relu', input_dim=10000))\n",
    "model.add(Dense(4, activation='relu'))\n",
    "model.add(Dense(1, activation='sigmoid'))\n",
    "model.compile(optimizer='rmsprop', loss='binary_crossentropy', metrics=['accuracy'])\n",
    "#\n",
    "small_history = model.fit(partial_x_train, partial_y_train, nb_epoch=20, batch_size=512, validation_data=(x_val, y_val))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 15000 samples, validate on 10000 samples\n",
      "Epoch 1/20\n",
      "15000/15000 [==============================] - 25s - loss: 0.5285 - acc: 0.7577 - val_loss: 0.2933 - val_acc: 0.8834\n",
      "Epoch 2/20\n",
      "15000/15000 [==============================] - 18s - loss: 0.2504 - acc: 0.9030 - val_loss: 0.6006 - val_acc: 0.7746\n",
      "Epoch 3/20\n",
      "15000/15000 [==============================] - 19s - loss: 0.1632 - acc: 0.9399 - val_loss: 0.4680 - val_acc: 0.8325\n",
      "Epoch 4/20\n",
      "15000/15000 [==============================] - 18s - loss: 0.1142 - acc: 0.9664 - val_loss: 0.3351 - val_acc: 0.8896\n",
      "Epoch 5/20\n",
      "15000/15000 [==============================] - 18s - loss: 0.0284 - acc: 0.9931 - val_loss: 0.9494 - val_acc: 0.7924\n",
      "Epoch 6/20\n",
      "15000/15000 [==============================] - 18s - loss: 0.0085 - acc: 0.9982 - val_loss: 0.5022 - val_acc: 0.8868\n",
      "Epoch 7/20\n",
      "15000/15000 [==============================] - 18s - loss: 0.0290 - acc: 0.9961 - val_loss: 1.4520 - val_acc: 0.7801\n",
      "Epoch 8/20\n",
      "15000/15000 [==============================] - 18s - loss: 0.0069 - acc: 0.9977 - val_loss: 0.5780 - val_acc: 0.8840\n",
      "Epoch 9/20\n",
      "15000/15000 [==============================] - 18s - loss: 1.6194e-04 - acc: 1.0000 - val_loss: 0.6645 - val_acc: 0.8838\n",
      "Epoch 10/20\n",
      "15000/15000 [==============================] - 21s - loss: 3.6084e-05 - acc: 1.0000 - val_loss: 0.7301 - val_acc: 0.8825\n",
      "Epoch 11/20\n",
      "15000/15000 [==============================] - 18s - loss: 9.4091e-06 - acc: 1.0000 - val_loss: 0.7934 - val_acc: 0.8813\n",
      "Epoch 12/20\n",
      "15000/15000 [==============================] - 18s - loss: 2.3008e-06 - acc: 1.0000 - val_loss: 0.8436 - val_acc: 0.8816\n",
      "Epoch 13/20\n",
      "15000/15000 [==============================] - 18s - loss: 6.5201e-07 - acc: 1.0000 - val_loss: 0.8896 - val_acc: 0.8813\n",
      "Epoch 14/20\n",
      "15000/15000 [==============================] - 18s - loss: 2.5566e-07 - acc: 1.0000 - val_loss: 0.9322 - val_acc: 0.8816\n",
      "Epoch 15/20\n",
      "15000/15000 [==============================] - 18s - loss: 1.4217e-07 - acc: 1.0000 - val_loss: 0.9652 - val_acc: 0.8814\n",
      "Epoch 16/20\n",
      "15000/15000 [==============================] - 18s - loss: 1.1897e-07 - acc: 1.0000 - val_loss: 0.9921 - val_acc: 0.8811\n",
      "Epoch 17/20\n",
      "15000/15000 [==============================] - 17s - loss: 1.1201e-07 - acc: 1.0000 - val_loss: 1.0123 - val_acc: 0.8814\n",
      "Epoch 18/20\n",
      "15000/15000 [==============================] - 17s - loss: 1.1042e-07 - acc: 1.0000 - val_loss: 1.0249 - val_acc: 0.8812\n",
      "Epoch 19/20\n",
      "15000/15000 [==============================] - 18s - loss: 1.0996e-07 - acc: 1.0000 - val_loss: 1.0319 - val_acc: 0.8811\n",
      "Epoch 20/20\n",
      "15000/15000 [==============================] - 18s - loss: 1.0981e-07 - acc: 1.0000 - val_loss: 1.0363 - val_acc: 0.8810\n"
     ]
    }
   ],
   "source": [
    "# 特徴ベクトルを大きく（512）にしたを使って分類する\n",
    "model = Sequential()\n",
    "model.add(Dense(512, activation='relu', input_dim=10000))\n",
    "model.add(Dense(512, activation='relu'))\n",
    "model.add(Dense(1, activation='sigmoid'))\n",
    "model.compile(optimizer='rmsprop', loss='binary_crossentropy', metrics=['accuracy'])\n",
    "#\n",
    "big_history = model.fit(partial_x_train, partial_y_train, nb_epoch=20, batch_size=512, validation_data=(x_val, y_val))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "org_val_loss = org_history.history['val_loss']\n",
    "small_val_loss = small_history.history['val_loss']\n",
    "big_val_loss = big_history.history['val_loss']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {
    "collapsed": false
   },
   "outputs": [],
   "source": [
    "dict = {'org_val_loss':org_val_loss,  'small_val_loss': small_val_loss,  'big_val_loss': big_val_loss}\n",
    "df_val_loss = pd.DataFrame(dict)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXcAAAEDCAYAAADOc0QpAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAIABJREFUeJzs3Xl8VNX5+PHPTDLZE7JNErIACUlOEggQ9h0BARcEBVSq\nVmu1rXWrbbWbv7Z+u1hb61arbd3qVneQTVYRlX1NICHkZF8h+74nM/f3RwADZpmQbSY579eLF8nc\ne+48c5M8c+bcc5+j0zQNRVEUZWjRD3YAiqIoSt9TyV1RFGUIUsldURRlCFLJXVEUZQhSyV1RFGUI\nUsldURRlCLK3ZCchxLPATMAMPCKlPNZuWzDwPmAATkgp7xdCLAA+BpIAHXBKSvmTvg5eURRF6Vi3\nyV0IMR8Il1LOFkJEAW8As9vt8gzwtJRykxDixfPJHuBLKeUtfR+yoiiK0h1LhmUWAxsApJQpgKcQ\nwg1ACKED5gKbz29/SEqZf76dru/DVRRFUSxhSXIPAErafV96/jEAI1ALPC+E2CuEeLLdfjFCiA1C\niK+FEFf3TbiKoiiKJSxJ7pf3wHWA1u7rIOA5YAEQJ4S4FkgFnpBS3gh8D3hdCGHR+L6iKIrSe5Yk\n3AK+6akDBAKF578uBbKllNkAQojdwDgp5TbaLqgipcwUQhTS9iaQ09mTaJqm6XRqJEdRFKU7mqbx\n1NtHOXDqHJufWdlh4rQkue8EngBeFULEAQVSyjoAKaVJCJEphBgrpcwApgDvCSFuA0ZKKZ8RQgQA\nfrS9SXRKp9NRUlJj+atTesRodFfntx+p89u/1Pm91M4juRw4dY7IEM9O9+l2WEZKeRA4LoTYD7wA\nPCCEuEsIsfL8Lj8F3hRC7AMqpZSbgU3AAiHE18CnwH1SytZevh5FUZRhLy2/ko+/zMDD1YH7Vo7r\ndD+dFZX81dQ7c/9RPZ/+pc5v/1Lnt011XTNP/PcIVXXN/OI7cYhRXhiN7h0Oy6g7VBVFUWyA2azx\nn02nqaxtZvWCsYhRXl3ur5K7oiiKDdiwL4szORVMCvflmhmjut1fJXdFURQrdyqjlC0HsvEd4cQ9\ny6PRWzCzUCV3RVEUK1Za1cCrm5Oxt9PzwE2xuDoZLGqnkruiKIqVamk18/KnSdQ1tnLH0khGB7hb\n3FYldyu0fHnn1RoKC89x7713DmA0iqIMlg92p5FdWMOc8QHMmzCyR21VcrdKXY+nqRt5FWXoO3i6\nkD3xBQQbXbljmaCnd/Crei9daG1t5W9/+zNnzxbQ2trKPff8kGee+RuzZ8/By8ub2bPn8ec//x53\ndw+EiKKyspLf/Ob33zqO2WzmlltW8v776zEYDMTHH2fdug/5yU8e5Y9//B06nY7W1lYef/wJAgOD\nLI7vxIljvPrqy9jbG/Dz8+PXv/49ZWWl/PGPv8POzg6TycRvf/sHAH760x9jNnPxMX//gG6OrijK\nYCkoqeWt7Sk4Odhx/02xOBrsenwMm0nuH32RztGU4j495rQoP25ZFN7p9s8/34GjoxP//OcrlJaW\n8tBDP8RkMjFr1hymTZvJ//t/v+D73/8hc+cu4He/+zVOTk4dHkev1zN16nSOHTvCrFlz2LfvaxYu\nvJqysjLuvvsHxMVN4bPPNvHpp5/wwAM/4Zu6bF175pmneOGFf+Hra+T5559m585t1NRUM23aDO66\n6x7S0iRlZaUkJp5kzpw5rFlzx8XHVHJXFOvU0NTKS58m0dxi5oGbxhPg7XJFx1HDMl1ISTlDXNwU\nAHx9fXFwcKCmpoqoqLZbfrOzs4mNnQjAnDnzujzW/PkL2b9/LwBHjhxk9ux5eHt78/HHH/Dggz/k\no4/ep6qq0uLYqqur0en0+PoaAYiLm0JaWirTp89i27bPeOmlF2hqaiYmZjzTps1kw4YNlzymKIr1\n0TSNN7elUFhez9JpIUwRfld8LJvpud+yKLzLXnZ/0Ol0tC/P0NLSgk6nx2C4MBVJuzgOptN1/T45\nbdoMXn75H2RmphMcPApnZ2eee+5vzJgxi5UrV/Hll7s5cGBfD2MzXxKbXq8jNDSMt956n6NHD/HK\nKy9x/fUrWLbsOjZt2sTWrbsueUxRFOuy+3g+R1OKCQ8ewZqrxvbqWKrn3oXo6Bji49uWiy0qKkSv\n1+Pm9s1UpKCgYFJSzgBw+PD+Lo9lMBgYOzac9957h4ULFwNQVVVJUFDbqoR7935Fa6vltdXc3d3R\n6fQUFxcBkJBwAiFi2L17J5mZ6cydu4B7772PlJQz7N69CynlJY8pimJdMgqq+PCLdNxdDPx45Xjs\n7XqXnm2m5z4YFi9eSnz8cR5++D5aW1t57LHH+dOfvrlgeued9/DXv/4Ro9Gf0NAw6urqujzeVVct\n4skn/4+f/ewXAKxcuZrnn3+agICRrF59K08//SRHjx7C0hUKf/GL3/DEE49jb29PYGAQV1+9lPT0\nVJ5++i+4uLhgZ2fHI488SmNjI3/84x8xGJyws9PzyCOPXvE5URSl79XUN/OvjUmYNY0frRiHl7tj\nr4+pqkL2wunTSTg7OxEWFs4777yJTgd33PG9wQ6rQ6qqXv9S57d/DeXzazZrPPfxSU5nlXPT/DBu\nmD2mR+07qwqpeu69YDDY85e//AFHRyecnJz4/e//zOOPP0ZNzTe/hJqm4ebmzl/+8vceHXvTpk/Z\ntWv7xTF9TWsb3//Rjx5k3Dh1QVRRhorNB7I5nVXOhLE+XD9rdJ8dV/Xch4mh3POxBur89q+hen6T\nMst47qOTeHs48fu7p+HmbFndmPZUPXdFURQrUl7dyCubk7Gz03H/TeOvKLF3RSV3RVGUAdZqMvPy\nhiRqG1r4zuIIQkd69PlzqOSuKIoygMyaxrs7U8k8W83Mcf5cFWd5yZGesOiCqhDiWWAmYAYekVIe\na7ctGHgfMAAnpJT3d9dGURRlOGpoauXVzckkpJcSZHTlrmVRPS4IZqlue+5CiPlAuJRyNnAv8I/L\ndnkGeFpKORMwCSGCLWijKIoyrBRXNvDkO8dJSC8lerQXv7xtMo4OPS8IZilLhmUWAxsApJQpgKcQ\nwg1ACKED5gKbz29/SEqZ31Wb4ejJJ/+Pgwf3sW3bFl566YU+Oaaq+a4otkPmVvCnt45RUFrH4snB\n/PSWiX1+AfVyliT3AKCk3fel5x8DMAK1wPNCiL1CiD9b0GZY67tPYKrmu6LYgi8TCvj7Bwk0NLVy\n5zLB7Usje11awBKWjLlfniZ0fFOTVgcEAc8BucAWIcR13bTplNHY+RJS7ySs41DeCQvCtdzMkMl8\nd9LqTrefO3eOxx577GJt9JkzZ1JQUEBFRQUZGRk88sgjbNmyhczMTJ5++mkmTJjAU089RWJiIk1N\nTaxdu5Y1a9bg5GRgxAgXTKZGnJ0dOnydZrOZq6++mu3bt+Pg4MCRI0d49913efzxx3nssccu1nx/\n6qmnCAkJQa/XdXq+mpursbe3w2h05/Dhwzz33HMYDAYCAgJ48sknKS0tveR1Pf300wDfemzkyJ6t\n/DLcdfX7q/SerZ1fk8nMaxuT2LI/C3cXB3591zRiw30H7PktSe4FXNrrDgQKz39dCmRLKbMBhBBf\nADFAfhdtOtXVTQr1Dc2YzH17w1V9Q3OXz7lu3UYmTZp6sTb64cMHycjI4qWXXmXz5g28/PK/+e9/\n/8dnn23ik08+xdc3GE9PX55//t80NTVx6603smDBMhobW6iqqqemppGGLp4zLm4q27d/waxZc9iy\nZTuzZy8gNTWHO+74/sWa76+//hYPPPATzGZzp8cpL6+jtdVESUkNv/3t73jhhX8RHR3G44//jvfe\n+5iamupLXldqajaJiSe/9Zi9/bAdSeuxoXqTjbWwtfNb29DCvzYkcSangiBfVx5aMwG/EY798ho6\ne9OzJLnvBJ4AXhVCxAEFUso6ACmlSQiRKYQYK6XMAKYA79GW9Dtsc6VWhS9nVfjy3hyix6ZNm8nj\njz9GbW0tCxYswsfHl6ioaAB8fHwZOzYcnU6Ht7cPiYknMRgMVFVV8eMffx97e0OP6rPDNzXfZ82a\nw5EjB7n33vuoqanm+eff4PXX/0NNTQ1CRFl8vI5qvickxLNixU38+tePXnxd48fH4uTkfMlrHT8+\ntkexK4rS5lxZHS98coriigYmhfvygxticHYc+Eov3Q78SCkPAseFEPuBF4AHhBB3CSFWnt/lp8Cb\nQoh9QKWUcnNHbfop/n4VFjaWN998n4kTJ/HKKy9RVFSInd03V7fbf61pGgkJJzhx4hgvvfQaL774\nHxwcHHr0fNOmzSAh4cQlNd9fe+3fzJgxi3/+8xXuvvveHh2vu5rvF17Xjh1bv/Vad+zY2qPnUhQF\nTmWU8ae3j1Fc0cB1M0fz4KrYQUnsYOE8dynlby57KLHdtgzgW8sQddDG5uzevZPAwCDmzl2Ah8cI\nHnvsJ6xcuarT/auqqvD390ev17Nv31eYTKYe1Wi3pOZ7T2oBta/5bjS6k5BwggkT4r71uvbs2Y29\nvYHAwMBLHlMLeiiKZTRNY+fRPD7ak46dXs8Plscwa/zgziFRVSG7EBIy6nxtdFfs7PT8+McPc/Zs\nfqf7T5s2nXfffZOHHvoR8+ZdxezZ83jmmad69Jz9VfPd2dkRozGgy5rv7V+rqvmuKJZpaTXz9o4U\n9icWMsLVgQdXxzI2cMRgh6WqQg4XtnZBytao89u/rPX8VtU189L6RNILqhgd4M5Dq2Lx9nAa0BhU\nPXcromq+K4rtyy2q4R/rTlFe3cT0aD/uvi4aR0P/3XHaU6rnPkxYa89nqFDnt39Z2/k9Lot5dUsy\nzS1mbpoXyvLZY/qtRkx3VM9dUZQBV1rVgO8I58EOo89omsbmA9ls2JuFo8GOB26KZYowDnZYHVLJ\nXVGUfpGQXso/PjnFzQvHcu2Mvls+bjC0tJo4dLqIXcfyyC+pw8fDkYdWT2CUv/XeNauSu6Io/SI+\nta281PqvMhkf6kOIn+3d8VxV18yeE/nsiS+gpr4FvU7HjBh/vrM4Ag/Xnt3HMtBUclcUpc9pmkZy\ndjn2dnpaTWZe3ZzMb++aisHeNtYHyi2qYdexPA4nF9Fq0nBxtOfaGaNYPCV4wGfDXCmV3BVF6XPF\nlQ2UVTcxVRhxcTLw9cmzbNqfxeoFYwc7tE6ZNY1T6WXsOpbHmZwKAPy9nFkyLYQ540f2a+31/qCS\nu6IofS45uy05xozxZkaMP8nZ5Ww9lMPEcF/Cgwb/Bp/2Gptb2Z9YyOfH8iiqaAAgerQXS6aFMGGs\nD3obrZ+tkruiKH0uOascgJgxXjg72nPP9dH87b14Xt+SzBN3T7eKXnB5dSO7j+fzVcJZ6ptasbfT\nMTd2JEumhdjk9YHLqeSuKEqfMps1zuRU4DvCCaNn2zRIMcqLpdND2HEkj0++zOD2pZGDFl/G2Sp2\nHc3jWEoJZk3D3cXAijljWDg5mBFWfpG0J1RyVxSlT+UU1VDf1MrUKOMlN/asmh9GYmY5u0/kMynS\nl3FjvAc0rvjUErYeziGjoBqAYKMrS6aFMDPGH4P94H+S6GsquSuK0qeSsy8MyVyavA32dty7PJo/\nv32cNz47wx/vmY6LU/+uI3rBtkM5fPxlBgATxvqwdFoI0aO9Bu2u0oFgG/OSFEWxGRcupkaP9vrW\ntjEBHtwwewwVNU2893nagMSz40guH3+ZgZe7I3+4ZzqP3DyRmDHeQzqxg0ruiqL0oaYWE2n5lYzy\nd8PdpePx6+tmjWZMgDsHkgo5Lkv6NZ5dR/P48It0PN0c+MVtcQQbbf9CqaVUclcUpc+k5VfSatK+\nNSTTnr2dnnuXx2Bvp+ftHSlU1zX3Syy7j+fz/u40Rrg68IvbJuPv5dIvz2OtVHJXFKXPfDO//dtD\nMu0F+rqyZkEYNfUtvLU9pUcrjFliT3wB/9uViodrW489wHt4JXZQyV1RlD7UVnJAR0SwZ7f7Xj0t\nBBHiSXxaKQeSCvsshq9PnuWdHRJ3FwOPrZ3ESB/XPju2LbFotowQ4llgJmAGHpFSHmu3LQvIPb9N\nA24HIoGPgSTa1oM7JaX8Sd+GriiKNampbya3qJaoUZ4WLVqh1+m45/pofvfGEd77PJXo0V69rtuy\n99RZ3tqWgpuzgcfWxhE0jMbYL9dtchdCzAfCpZSzhRBRwBvA7Ha7aMA1UsqGdm0igS+llLf0dcCK\nolinC/VYxoVaPn/d19OZtYsjeHNbCq9/doafr510xbf7H0g6x5tbU3BxsufRtZMIHgJ3mfaGJcMy\ni4ENAFLKFMBTCNH+rOnoeLXmoT3PSFGUS7SvJ9MT8yaMZMJYH87kVLDnRMEVPfeh04W8/tkZnB3t\neXRtnFXXWR8oliT3AKD9fKXS84+1928hxF4hxJPtHosRQmwQQnwthLi6t4EqimLdkrPLcXG0Z3QP\nE6tOp+Pua6Nwczbw8Z50Csvre9T+yJkiXt2SjJODPT9fO4nRASqxg2Vj7pf3wHW0DcVc8FtgO1AO\nbBRCrAIOAk9IKT8WQoQBe4QQY6WUrV09kdGofij9SZ3f/jWcz++50jpKqxqZFTsSf3+PHrc3Gt15\n4OaJ/PXtY7y1Q/LXB+ZiZ6f/1j6X23/yLK9sTsbZ0Z4//mg2kaO6nqUznFiS3Au4tKceCFy8tC2l\nfPfC10KIrUCslHI9bRdUkVJmCiEKgSAgp6snsqYFcIcaa1tgeKgZ7ud3X3zbcEr4yCs/DyLQgxkx\n/hxOLuLtLadZPnvMxW0dnd/jsoR/b0zCYK/nkZsn4uVsPyx/Bp11KiwZltkJrAEQQsQBBVLKuvPf\newghtgshLhSIWAAkCSFuE0L8/Pw+AYAfbW8SiqIMQZ3Vk+mp25dEMsLNgY37ssgt6jxRx6e1JXZ7\nOz0/vXmi1dWItwbdJncp5UHguBBiP/AC8IAQ4i4hxEopZTXwGXBICLEXKJZSrgM2AQuEEF8DnwL3\ndTckoyiKbbpQ4tfHwxE/L+deHcvN2cD3r4vGZNZ4bUsyLa3mb+1zMr2Ulz9Nws5OxyM3TyAypPs5\n9cORRfPcpZS/ueyhxHbbXgRevGz/WmBFr6NTFMXq5RbXUNfYyuRIY58U44oN8+GqSYF8mXCWjfuy\nWHPVN0vzJWWW8dKnidjpdTyyZiJCjbF3St2hqihKr1zpFMiu3LIoHKOnE9sO55CeXwXA6exy/rEu\nEZ1Ox8NrJhDVQdVJ5RsquSuK0isXxts7KvF7pZwc7Lnn+hjQ4LUtyRw5Xcg/PjkFwEOrY/v0jWSo\nUsldUZQr1txiIjWvihA/Nzz6eIm6yBBPls0YRXFlA3984zCapvHgqvGMD/Xp0+cZqlRyVxTliqUV\nVNFqMndbBfJK3TQvlGCjK/Z2Ou6/KZYJY3375XmGIrXMnqIoV6yvpkB2xmBvx6/vmIKTiyM6k6lf\nnmOoUj13RVGuWHJ2BfZ2OiItKPF7pZwd7fEbhvXYe0sld0VRrkhtQwu5hTWEB43A0aH7Er/KwFLJ\nXVGUK5KSU4EGRKuZK1ZJJXdFUa7I6Yvj7Wq+uTVSyV1RlCuSnF2Os6M9Y1SJXaukkruiKD1WXNlA\nSWUjUaM8sdOrNGKN1E9FUZQeO9PPUyCV3lPJXVGUHvumnowab7dWKrkritIjZq2txK+3hyMBav65\n1VLJXVGUHskrqqW2oYWY0d59UuJX6R8quSuK0iPJagqkTVDJXVGUHrlY4lddTLVqKrkrimKxllYT\nqflVBBtdGdHHJX6VvqWSu6IoFkvPr6Kl1aymQNoAi0r+CiGeBWYCZuARKeWxdtuygNzz2zTgdinl\nua7aKIpim5Jz1BRIW9FtchdCzAfCpZSzhRBRwBvA7Ha7aMA1UsqGHrRRFMUGJWeXY6fXERnSfyV+\nlb5hybDMYmADgJQyBfAUQri12647/68nbRRFsTG1DS1kn6thbNAInBzUOj/WzpLkHgCUtPu+9Pxj\n7f1bCLFXCPFkD9ooimJDLpT4VUMytsGSt9/Le+U62oZiLvgtsB0oBzYIIVZb0KZDRqOqLtef1Pnt\nX0P9/GZ9lQnAnEnBg/Jah/r57WuWJPcCLu11BwKFF76RUr574WshxDYgFsjvqk1nSkpqLAhHuRJG\no7s6v/1oOJzf4ylFODnY4elsN+CvdTic3yvV2ZueJcMyO4E1AEKIOKBASll3/nsPIcR2IYTh/L4L\ngERgV2dtFEWxPaWVDRRXNBA1ykuV+LUR3fbcpZQHhRDHhRD7ARPwgBDiLqBSSrlRCPEZcEgIUQ/E\nSynXAVzeph9fg6Io/UxNgbQ9Fl3yllL+5rKHEtttexF40YI2iqLYqAslB8aFqpuXbIX6fKUoSpfM\nmkZydgVe7qrEry1RyV1RlC7lF18o8eulSvzaEJXcFUXp0jerLqkhGVuikruiKF36psSvuphqS1Ry\nVxSlUy2tZlLzKgnydcXTzXGww1F6QCV3RVE6lVFQRXOrWfXabZBK7oqidOr0xSX11Hi7rVHJXVGU\nTiVnV2Cn1yFUiV+bo5K7oigdqmtsIbuwmrBAD5wdVYlfW6OSu6IoHUrJqUTT1JCMrVLJXVGUDiXn\nXBhvVxdTbZFK7oqidCg5uwJHBztCR3oMdijKFVDJXVGUbymraqSovJ6oEE/s7VSasEXqp6Yoyrck\nqymQNk8ld0VRvuVi/XZV4tdmWU1yzyzPGewQFGXYa2oxsetoHgnppYxwcyDQR5X4tVVWM3n130ff\n5bHJDw92GIoyLDU1m9gTX8D2I7lU1zXjaLBjzYKxqsSvDbOa5J5dmU9pQxm+zj6DHYqiDBsNTa1t\nSf1wLrUNLTg52LF89miWTA3B3cVhsMNTesFqkjtAfHEiS0ZfNdhhKMqQV9/Yyu4T+ew8kktdYyvO\njvasmDOGJdNCcHUydH8AxepZlNyFEM8CMwEz8IiU8lgH+/wFmCmlXCiEWAB8DCQBOuCUlPInXT2H\nXqcnvkQld0XpT3WNLXx+LJ9dR/Oob2rF1cmem+aFsnhKCC5OVtXXU3qp25+mEGI+EC6lnC2EiALe\nAGZftk80MA9obvfwl1LKWywNZJxfJIlFKZQ3VuDtpO6IU5S+VNvQws6jeew+nkdDkwk3ZwOrF4Sx\naHKwqhszRFkyW2YxsAFASpkCeAoh3C7b5xngN5c91qMrMTODJwOQUJLUk2aKonShur6ZT77M4LF/\nHWDLgWwMdnpuWRjO3348i+tnjVGJfQiz5CcbALQfhik9/1g6gBDiLmAPcPlcxhghxAbAG/iDlPLz\nrp5kWvBEXjv+PvHFiSwKmWdp/IqidKCqrpkdh3P5Ij6f5hYzI1wduGleGAsmBeJosBvs8JQ+UFB7\nji/y9vIz4z0dbrckuV/eA9cBGoAQwgu4m7befUi7fdOAJ6SUHwshwoA9QoixUsrWzp7E08mDaGM4\nySVp2LmZ8HZW9aP7mtHoPtghDGnWcH6rapv4aHcq2w/m0NxiwmeEE2sWRbBkxmibT+rWcH6tQXNr\nM58kb2Vzyi5Mmhm48uReQFtP/YJAoPD814sAX2Av4ASECSGekVL+nLYLqkgpM4UQhUAQ3+7dX2Kc\nVwzJJWl8kXKYBcGzu9pV6SGj0Z2SkprBDmPIsobzm5RVxmtbzlBd14yPhyPXLQpnbuxIDPZ6qivr\nBzW23rKG82sNUsrTeF+up7ShDB8nL24Vqzrd15LkvhN4AnhVCBEHFEgp6wCklOuAdQBCiNHAf6WU\nPxdC3AaMlFI+I4QIAPxoe5Po0iTjeD5O3Uh88SmV3BXFQq0mM+u/zmT74Vzs9DpuXjiWJVNDVMGv\nIaS2pY71aVs4XHgcHToWh8zn+rClONp1fi9Ct8ldSnlQCHFcCLEfMAEPnB9nr5RSbuyk2SbgPSHE\nSsAA3NfVkMwFno4jCBsxmvTKLGqaa3F3uPy6raIo7RVV1POfjafJLqzBz8uZ+1aOY0yAKtE7VGia\nxrGiBD5J20RtSx0hboHcFr2GUe7B3bbVaZo2ACFaRCspqeGL3K9Zl76F74hVzA2aOdgxDRnqY23/\nGozze/B0IW/vkDQ1m5gzPoDblkQO2dkvw/H3t7ShnA/kes6Up2LQG1getpSFwXOx01967cRodO9w\nZqLV/SZMNMayLn0L8cWJKrkrSgcamlr5365UDiQV4uhgxw9uiGHWuIDuGyo2wWQ2sSd/H59l7qTZ\n3EK0dyRrxSp8nXtWodPqkruPsxej3UNIrcygtqUON4PrYIekKFYju7Caf288TXFFA6Ej3fnRinH4\neanKjUNFbk0+76WsI6+mADeDK9+JWs00/7grKuBmdckdIM4vlpyaPBJLkpkVOG2ww1GUQWfWNHYe\nyWPdVxmYzBrXzhjFTfPD1EXTIaLJ1MxnWTvZk7cPs2ZmRsAUVoUvx83hyju3VpncJxlj2ZCxlfiS\nRJXclWGvqq6Z1z9LJimzHA9XB+5dHs34UFU9dahILpN8INdT1liBr5M334laTZR3RK+Pa5XJ3eji\nQ7BbICnladS3NOBicB7skBRlULSfuz4+zJt7r4/Bw1WV4h0KapprWZe2maNF8eh1epaMuorrQq/G\noYvpjT1hlckd2oZm8jPPklR2hukBkwc7HEUZUJfPXb91UThLpoWgV4tn2LSG1gbyas6SXZXL57lf\nUddaz2j3EG6LWk2we2CfPpfVJvdJxlg2Z+4gvjhRJXdlWFFz14eGupZ68moKLv7LrcmnpKHs4nYH\nOwdWR9zAVcFz0Ov6/tqJ1Sb3AFc/Rrr6k1wuaWxtxMneabBDUpR+dzCpkLd3Do+560NJbXPdxQTe\n9n8BZY3ll+zjYu+M8ApnlHswIe6BRHiNxcOh/+rlWPVvTZwxlq3Zn3O6LIUp/pMGOxxF6TctrSbe\n2i7V3HUbUN1cQ251Pnk1Z8mrySe3poCKpspL9nE1uBDtHUmIe9D5ZB6Ej5PXgK5Ja9XJfZJfW3KP\nL05UyV0ZshqbW3lxXSJncirU3HUrZtbMrE/fwp68fZc87u7gxjifqPOJPIgQ9yC8HD0HfXFxq07u\nga4B+Ln4croshWZTc59dRVYUa1Hf2MJzH58ko6CauAhf7ls5HoO9mrtubVpMLbyZ/AEJJYn4ufgy\n1W8SozwDifWAAAAgAElEQVTaeuQjHDwGPZF3xKqTu06nI844gR05X5BcJpnkFzvYISlKn6mub+bZ\nDxLILa5l5jh/7rk+Gju9SuzWpq6lnv+cepOMqmwiPMP4YeyduBis/5OV1f8mTfIbD0B8SeIgR6Io\nfaeipom//u8EucW1LJgUyL3LY1Rit0JlDRU8e/xlMqqymew3gQcm3WsTiR2svOcOEOIWhI+TN4ml\nybSYWjDYGQY7JEXplZLKBp5+P57SqkaWTgvh1kXhVvmxfrjLqznLyydfp7q5hkUh87gp/Pp+mbLY\nX6w+Up1OR5xfLE2mZlIq0gY7HEXplXNldTz1vxOUVjWyYs4Yldit1JnyVJ478TI1zbWsjriB1RE3\n2FRiBxtI7tB2QxNAfLEamlFsV25RDU/97wQVNU3csjCcG+eFqcRuhQ6fO87LJ9/ApJn5/vjbWRQy\nb7BDuiJWPywDMMYjBC9HT06VnqbV3Iq93ibCVpSLMgqqeO6jkzQ0tfLdZYKFcUGDHZJyGU3T2Jmz\nh02Z23G2d+a+Cd8j3DN0sMO6YjbRc9fpdEzyG09DayOyImOww1GUHknJqeDvHyTQ0NzKPcujVWK3\nQmbNzIepG9iUuR0vR09+PuV+m07sYCPJHb4ZmkkoPjXIkSiK5U5llPLcxydpNZm5/8bxzB4/crBD\nUi7TbGrmlcS32VtwkCC3kTw69QFGuvoPdli9ZtH4hhDiWWAmYAYekVIe62CfvwAzpZQLLW3TE2Ej\nRjPCwZ2TpadZa171rXUEFcXaHEsp5j+bTqPX63h4zQRiw1QNdmtT21zHv0/9l6zqXIRXOD+IvRPn\nIVLHqtueuxBiPhAupZwN3Av8o4N9ooF5gGZpmx4HqtMz0RhLXUs9aZWZvT2covSr/Ynn+NfGJOzt\n9fzslokqsVuhkvoynjn+ElnVuUzzn8z9E78/ZBI7WDYssxjYACClTAE8hRBul+3zDPCbHrbpsbjz\nd6iqG5oUa/bFiXxe/+wMLo72PLY2DjHKa7BDUi6TU53HM8dforihlKWjF3JXzK1DbqKGJck9AChp\n933p+ccAEELcBewBcixtc6XCPUNxM7hysjgJs2bu7eEUpc9tPZTDuztT8XAx8IvbJhMWqOqwW5vT\nZSk8H/8falvquDXyRlaOvXZITkm15K3q8let45vhFy/gbtp66iGWtOmK0dh9beOZIXF8nrmPMoqJ\nMfZ+ncHhxJLzq1wZTdPYcSyfT77MwHeEE3/68RyCjL3+sKq00xe/v19k7ueVU+9hp7fj0bk/YlrQ\nxD6IzDpZktwLuLTXHQgUnv96EeAL7AWcgDAhxDNAPjCykzadKimp6TaYKI8oPmcfe9IOY9SpeteW\nMhrdLTq/Ss9pmsbGAzls2puJn6czj66dhAOaOt99qLe/v63mVrZn72Zb9m5cDS7cN+FuxjiMHhI/\no87e9CwZltkJrAEQQsQBBVLKOgAp5Top5fjzF05vAk5IKX8O7AJWd9SmtyI9x+Ji70xCcaIamlEG\nXavJzH+3prBpbyaBvq788vbJ+HqqBd2thVkzc6TwBH889He2Ze/Gx8mbn0++n7ARowc7tH7Xbc9d\nSnlQCHFcCLEfMAEPnB9nr5RSbrS0TV8FbKe3Y4JxHIfOHSO7Om9Y/JAU61RT38xLnyaRmldJePAI\nHloVi7uLWnPAGmiaRmJpMpszd3C2rhB7nR1XBc/h2jFX4+bgOtjhDQidpnU7FD5QNEs/IiWVnuFf\np/7LopB5rI64oZ/DGhrUsEzfyi+p5R+fnKK0qpEpwsiv7ppOTXXDYIc1ZPXk9ze1Ip1NGdvJqs5F\nh44ZAVO4LnQJPs5Dc9aS0eje4dVgm5z7I7wjcLJzIqEkiVXhy4fklW7FeiWklfKfzadpajaxYs4Y\nVswNxcnRHvXWObhyqvPYlLH9YvXYScbxLA9bNiTuNr0SVpPcy6os7/UY9PbE+sZwtOgEuTX5jPYI\n6b6RovSSpmlsP5zLJ19mYG+v576V45gePTwThzUprCtic+YOEkqSAIjyimDF2GuGfV6wmuT+f68d\n4vHvTkFvYS88zi+Wo0UniC9OHPY/RKX/tbSaeGu75EBSIZ5uDjy0egKhI9Uc9sFU1lDOZ1m7OFJ4\nAg2NMR6jWDn2GiK9wgc7NKtgNck962w1R88UMyPGsp5QtHckDnYOxJckDtmbEBTrUFXbxD8/TSSj\noJrQke48uGoCXu6Ogx3WsFXdXMP27C/YV3AIk2Yi0DWAG8KWEesbo/JAO1aT3O3tdHy6N5Mpwoi9\nXfczNB3sDMT6RHO8+CT5tecIcQ8cgCiV4SansIYX15+ivLqJmTH+fO/aKBwMqmjdYKhvaeDz3K/Y\nk7eXZnMLPk7eLA9bylT/STa3StJAsJrkvnTGaLYeyGZ/4jkWTLKs3vUkv1iOF58koSRRJXelzx1L\nKea1z5JpbjGzekEY180crXqGg6DZ1MyGMzvYkLyD+tYGPBzcuWnMcmYHThty9WD6ktWcmVuXCD4/\nksum/dnMGhdgUe9onE8UBr2B+OJTLA9dqv7wlD6haRqbD2SzYW8WjgY7HlwVy+RI42CHNexomkZ8\nSSLr07ZQ0VSJi70zK8dey1XBc3CwU/cTdMdqkru3hxOLpwaz7VAuX5wo4JoZo7pt42jnwDgfQUJJ\nEufqigh0U+UIlN5pbjHxxtYzHDlTjI+HIw+tnsAof1WTZ6AV1hXxceomUirSsNfZcWP0Mub4zsbF\noO7+tZTVJHeAa2eM5sv4s2w9lMOCSYE4O3YfXpwxloSSJBJKElVyV3qloqaJF9edIruwhvDgETx4\nUywerqqHOJAaWxvZlr2bL/L2YtbMxHgLbo5cwbjRYeomvB6yquTu5mzgmhmj+PTrTHYcyeXGeWHd\nthnnG429zo744kSuC10yAFEqQ1HWuWr+se4UVbXNzIkN4M5lURjs1UW6gaJpGseLElif/hlVzdX4\nOHmxOmIFE9QMmCtmVckdYMnUYHYfy2PH0TwWTQnGo5taHc72TkT7RJJYeoai+hL8XdTYqNIzh5IL\n+e/WFFpbzdyyMJxl00NUQhlAZ2sL+Sh1A2mVmdjr7bluzNUsGb0QBzvDYIdm06yua+LkYM/y2WNo\najax9WBO9w2AOOMEABKK1QpNiuXMZo31X2fyyqZk7PQ6fnLzBK6ZMUol9gHS0NrAJ6mb+MvR50mr\nzCTWN5rfzvg514ctVYm9D1hdzx1gwaQgdhzJ44sTBSydFoK3R9frGsb6RqPX6YkvSWTZmEUDFKVi\nizRNI7eoloOnCzl8poiq2maMnk48vGYiQb7Do1rgYNM0jSOFJ/g04zNqmmvxdfbh5ogVjPeNHuzQ\nhhSrTO4Gez0r54byxtYzbNqfzfeujepyfxeDC1FeESSXS0rqyzC6qMWIlUuVVjZwKLmIQ8lFnC1t\nW1rA1cmeq+KCWDU/DDdn1VMcCHk1Z/kodQOZVdkY9AZuCFvG4pD5GFRPvc9ZZXIHmDXen22Hc9h3\n6hzXzBhFgLdLl/tPC4gjuVyyLn0zP4q9S320VqhtaOFYSjEHTxeSll8FgL2dnqnCyMxxAcSG+aiL\npgOkvqWezZk72VtwEA2NScZYVoUvH7JleK2B1SZ3O72em+aF8fKGJDbszeS+leO73H+q/yQOnj1K\nYmkyRwpPMGPklAGKVLEmzS0mTmaUceh0IacyyjCZNXRA1ChPZo0LYIow4uKkeokDxayZOXTuGBsz\ntlHbUoe/i5GbI1YS7RM52KENeVab3AGmCCOjA9w5cqaY62bWdHkziV6n547om/nzkWf5OG0jkV5j\n8XLyHMBolcFiNmvI3AoOni7ieGoxDU0mAIKNbswa78+MaP9ur9sofcesmSmoLSS9MpOjhfHk1OTh\nYOfAjWOvY2HIXFUyYIBY9VnW6XSsXhDGsx+eZP3XmTxyc9crlfs4e7M64gbeS1nH/1I+4YGJ96jh\nmSFK0zTyims5dLqIQ8mFVNY2A+Dt4chVcUHMigkg2M9tkKMcHsyamfzas6RVZJJWmUlGZRb1rd+s\nzzDFbyI3hV+vOlsDzKqTO8C4Md5EjfLkVEYZqXmVRIZ0/Qsye+R0EkqSSC6T7D97mLlBMwcoUmWg\nyNwK3t+dRm5RLQAujvbMnxjIrHH+RIR4WrwmgHJlTGYT+bVnSa3IIL0yk4yqbBpaGy9u93HyZoLv\nOMK9woj0DMPH2XsQox2+LEruQohngZmAGXhESnms3bYfAN8HWoGTUsoHhRALgI+BJEAHnJJS/uRK\nAtTpdKxaMJYn3znO+q8y+OXtk7vsjet0Om6PWsOfDj/LuvQtRHlH4OusZs8MBdV1zXy0J50DSYXo\ngMmRRmaNC2DCWHVhtD+ZzCZya/JJq8wkrSKTzKpsGk1NF7cbnX2IM8YS7hlGhFcY3k7qIqk16Da5\nCyHmA+FSytlCiCjgDWD2+W3OwC3AHCmlWQixWwhxoav8pZTylr4IMjxoBJPCfUlILyUxs5wJY7tO\n1p6OI7glciVvJX/Au2c+5uG4H6p6zzbMbNb4KqGAdV9lUt/Uymh/d767TBAWqFZC6g+appFXU0By\neSppFRlkVufQbGq+uN3fxcgUzzAizidzT8cRgxit0hlLeu6LgQ0AUsoUIYSnEMJNSlkrpWwAlgAI\nIVwAD6AQGE1bj73P3DQ/jJPppaz/OoPxYd7dfvSe5h9HQkkSJ0uS+Cr/AAtD5vZlOMoAyS6s5p0d\nkqxzNTg72nH7kkgWxgWh16uhl77W2NrE8aIE9hYcJK/27MXHA1z92xK5ZyjhnmGMcFRvqrbAkuQe\nABxr933p+cfSLzwghPgl8DDwvJQyWwgxGogRQmwAvIE/SCk/702gIX5uzIjx51ByEcdSirtdmFin\n0/EdsYqMyiw2ZmwlxjsSf1e/3oSgDKD6xhY+/TqLL+Lz0TSYGePPrYvCGeGmlrfrawW159hXcIgj\nhSdoNDWh1+mZ6DuOqQFxRHiG4e6gLkzbIp2maV3uIIR4Bdgspdx8/vu9wPeklBmX7ecIbAMeB7KB\nuVLKj4UQYcAeYKyUsrWLp+o6EOBsaS33//ULAnxceOmxRdhZsBzfobwTPHvgVSJ8QvnDop9jp1dL\npFkzTdP46kQ+r28+TWVNE0FGN368egITI1RBuL7UbGrhUN4JdmXsRZa2/Sl7O3uyOGwOi8Lm4OOi\nxs1tSIcfYy3puRfQ1lO/IJC2oReEEF7AeCnlXillkxBiG23j7wdpu6CKlDJTCFEIBAFdVgLrrl6z\nAZg3YSRfJpxlw5405k/sfmm9sU4RTPWfxLGiBD44/hlLxyzsts1QZDS6W3097LOldby7U5KSW4mD\nvZ7VC8JYOm0UBnu91cduC+cXoLi+lH1nD3Ho3DHqWuqBtsXm5wXNZLxPNHZ6O8x1UFJnXa/FVs7v\nYDAaO77/x5LkvhN4AnhVCBEHFEgp685vMwBvCiFipZT1wHTgbSHEbcBIKeUzQogAwI+2N4leu2FO\nKPuTCtm4L4tZ4/wx2HffE78l8kbSKjLYkrWTcb5RBLmN7ItQlD7S1GJiy4Fsth/OxWTWmDjWh9uW\nRGL0VKvu9AWT2URiaTJ7Cw6RUpEGgJvBlSWjrmJO4AxVi2mI6nZYBkAI8SSwADABDwCTgUop5UYh\nxJ3Ag0ALbVMh7xdCuAHvAZ60vQE8IaXc0c3TaJa+M3+0J53th3NZuziCpdNCLGqTVHqGf536LyFu\ngTw69cFhd5ectfZ84tNKeG9XGmXVjfh4OHLb1ZHE2eB6pdZ4fisaK9l/9ggHzh6hqrkagLEjQpkX\nNJNJfrEYbOhvwBrPr7UwGt07HJaxKLkPEIuTe21DC7/89wHs9Hr+et8si5bjA/jfmY85cO4o1465\nmuVhS3sTq82xtj+O0soG3vs8jYT0Uuz0OpZNH8UNs8fg6GCb10Ss5fyaNTNnytPYV3CIxNJkNDSc\n7JyYMXIycwNn2uxSlNZyfq1RZ8nddt6623FzNrBs+ig27M1i19E8VswNtajdqogbOFOexo6cL5jg\nG8Moj+B+jlS5XKvJzI4juWzen01zq5moUZ7cvlSoWup9oLShjLeTPySjKhuAEPcg5gXNZKp/HI52\nai3Y4cYmkzvAkqkh7D6ez/YjuSyaEmxRPW5neyfuiL6ZFxNe5a0zH/KrqQ+rOtIDqKnFxHMfnSQ1\nrxIPFwN3XRPFzHH+qv5PL2maxoFzR1iXtpkmUzMTfcexbMwiRntYNmSpDE02e9ums6M9188aQ2MP\nluMDiPKOYH7QbArritiStbMfI1Taa2k189L6RFLzKpkcaeTPP5zJrPEBKrH3UlVTDf8+9SbvpaxD\nr9NzV8xafhB7p0rsiu0md4CFcYF4eziy+0Q+FTVN3Tc478bw6/B19mF37tdknv8Iq/Qfk9nMK5tP\nk5TVVjrivpXjcFU11XstvjiRPx95hqSyMwivcB6f/jOmB3Rde0kZPmw6uRvs7VgxJ5SWVjOb92dZ\n3M7RzoE7o28F4O3kD2lqVzdD6VtmTePNrSkclyWIEE/uv3E89hbcfKZ0rr6lgbeSP+C1pHdoNjVz\nc8RKHpx0ryqpq1zC5v/K5sQG4O/twt5T5yiqqLe43VjPMSwaNY+ShjI2ZmzrxwiHL03TeH9XGvuT\nCgkd6cHDaybgYLDN2TDWIqU8jSePPMeRwhOMdg/hV9Me4aqQOaownvItNv8b0bYcXygms8bGvZb3\n3gFuCF1GgIsfX+XvR5and99A6ZH1X2ey+0Q+QUZXfnrLRIunrCrf1mxq4ZPUTbyY8CpVzdVcF7qE\nn0+5nwBVL0nphM0nd4CpUX6M8nfjcHIROYWWz4U12Bm4M+ZW9Do975z56JIFB5Te+exgNp8dzMHP\ny5lHb51k0WwmpWM51Xk8dfQF9uTvw9/Fj0enPMD1oUtUnSSlS0Miuet1Om5eGI4GvL0jBbPZ8huz\nRnuEsGz0QiqaKlmftqX/ghxGvjiRz7qvMvH2cOTRtZNUJccrZDKb+CxrF38//hJF9cUsDJ7Lr6b9\nRM2EUSwyJJI7tC3HN3OcP1nnath9PL9Hba8Zs5hgt0AOnDtCUumZfopweNifeI53d6bi4WLg0bVx\n+I5Q9WGuRGFdMc8cf5mtWbvwcHDnoUk/YE3kChzUfRmKhYZMcgdYuygCVyd71n+dSWlVQ/cNzrPX\n23NnzK3Y6ex4L+WTi9XylJ45Lot5Y+sZXBzt+fnaOAK8XQY7JJtj1sx8mbefp44+T05NHtMDJvP4\n9J8R5R0x2KEpNmZIJXcPVwfWLo6gqcXEuztT6UndnCC3kVwXuoSq5hreS/mEFlNLP0Y69CRllvHv\njadxMNjx01snEuKnFnjoqYrGSl5KeJ2P0zbiYOfAveO/y10xa3ExqE8/Ss8NuekLs8cHcCCpkFMZ\nZRy1YMWm9paMWsDpsjMklCTx9PF/8r2Y79hsoaWBlJpXyT/XJ6LX6/jJ6gmMDVRralpC0zSKG0pJ\nKU9r+1eRRrOpmfE+0dwWtYYRjh3X6VYUS9hkVcjuFFXU87vXj+DsYMeffzizR3dDNpmaWZ+2mX1n\nD2Ovt+emsdezIHi2zd/1119V9bILq3n6/XiaW8w8uCqWieG+ff4ctsDS81vTXIusSL+Y0CuaKi9u\n83X2Yenoq5g9crrN/771NVUVsnNDquSvJbYeyuGTLzOYN2Ekd18X3eP2J0tO817KJ9S21BHjLbgj\n+hab7kn1xx9HQWkdf/3fCeoaW/jRinE9+pQ01HR2fptNLWRUZV1M5vntFp52tXch0jucKK9worwj\n8XX2HsiQbYpK7p0bUiV/LbF0WgiHk4vYe+ocs8YFEDW6Z2tCTjSOY4xHCO+c+YjkcsmTR57ljuib\nifWN6aeIbUtxZQN//yCe2oYW7r42algn9vbMmpn82rMXk3lGVTat5ralg+11dkR6jiXKO4Io7whC\n3IPUnaVKvxmyPXeArHPV/OmtY/h5OfOHe6ZbtCTf5cyama/yD7AhYyut5lbmBc1iVfj1ONhYfey+\n7PmUVzfy1P9OUFrV2KPVsIaqqqYacpuzOJqTiKxIp7al7uK2ILeRRHm1JfNwz1Cb+72xFqrn3rlh\n13MHCB3pweKpwXx+LJ/NB7JZNX9sj4+h1+lZGDKXSK+xvHn6ffYWHCS1IoO7x32HEPegfojaulXX\nNfPMhwmUVjVy47zQYZvYNU0jqzqXr/L3c6L4FGbNDICn4whmBkwlyjsC4R2Oh4PtDuUptm1I99wB\nGptb+e1rh6msbeb3d08j2HjlU/RaTC1szNjGnvx92OnsuCFsGYtHzbeJj9Z90fOpb2zhb+/Fk1tc\nyzXTR3HzwrHD7sJfi7mVE0Un+TJ/P7k1bTfLjXT1Z2nEPEY5jsbfxW/YnZOBoHrunevVBVUhxLPA\nTMAMPCKlPNZu2w+A7wOttC2Q/WB3bTrRL8kd4GR6KS98coqxgR78+o4p6PW9++NLLpO8c+Yjqptr\niPQK587oW6y+3OqV/nFomkZpVSP5xbVsPZRDxtlqFkwK5M5lYlglscqmKvYVHGJfwWFqWmrRoWOC\nbwwLgucQ6TUWPz8PlXz6kUrunbviYRkhxHwgXEo5WwgRBbwBzD6/zRm4BZgjpTQLIXYLIWYCDp21\nGQwTw32ZHu3HkTPF7IkvYPGU3q2dGuMj+M30n/K/lE9ILE3mySPP8Z2o1Uz2m9BHEQ+OhqZWCkrr\nyCuuJb+4lryStv8bm00X95kZ4893lw5uYk+ryGB9+mfodDrCPUOJ8AwjbMQYXA19e0fshaGXL/P2\nEV+SiFkz42zvzNWjFjA/aBY+anaLYsW67bkLIf4PyJFSvnH++2RgupSy9rL9XICvgJuBuy1pc5l+\n67kDVNU28firhzFrGn+6dwbeHk69Pqamaew7e5h1aZtpMbcwc+RUbo5YgZN974/d19r3fMyaRkll\nQ1sCP/8vv6SWkspLq2LqdToCfFwINroS4ufGaH93YsZ49/qTz5VqNjWzKWM7e/L3oUOHTqe7ONYN\nEOgaQLhnGOGeoYR7hjLC0eOKnqejoZdA1wAWBM9mWsDkDhebVj3L/qXOb+d6c0E1AGg/pFJ6/rGL\nBdCFEL8EHgael1JmCyG6bTPQRrg5csuicN7clsK7O1N5aHVsr3ufOp2OeUEzifAM483k9zl07hjp\nlVl8L2YtoSNG91HkvVdQWseR1FLOZJaSX1xLfkkdTS2mS/ZxczYQPdqLED83go1uhPi5EejrckUz\njPpDZlUO7yR/SHFDKf4uRr4bfStBbgFkV+eSVplFemUWWVU5nK0r5OuCAwAYnX0uSfY+Tt5d/sw7\nGnqZ6DuOq0LmEOE5/K4vKLbNkuR++W+0Drikuy+l/KsQ4nlgmxBivyVtOmI09u/MglWLIzmWWkJC\neilphbXMmRDYJ8c1Gt35a8iv+DBpM5tSdvHsiX+xZtz13BS9bFBrblfVNvHOtjPsPJzDhQ9odnod\nIf7ujAn0IHSkB2NGjmBMoAde7o5WmbyaTS18lLSFzXIXaLA8cjFrY1fgYN/Wew4K8GEOcQC0mlrJ\nrMjlTEk6Z0rSSCnN4OC5oxw8dxQAb2dPoo3h5/9FEOQRgA4dqWWZbEv7ksN5JzBpZlwdXFgRtYSl\n4Qvwc/WxONb+/v0d7tT57RlLhmV+D5yVUr56/vsMYIKUsk4I4QWMl1LuPb/tsfPNnIFzHbXp4qn6\ndVjmgsLyttIErs72/PneGbj08ULNqRXpvJX8IZVNVQS4+HFD2DImGscPaOJsNZnZE1/Ahr1ZNDS1\nEuTryprFkXi7Ghjp42Iza5jmVOfx9pmPKKwrwtfZh+9G30K4Z6jF7c2amYLaQjIqs0ivzCS9Moua\nlm9GBl0NLrg7uFNYVwS0Db1cFTyHaQFxPZ6ProYN+pc6v5274tkyQohZwBNSymVCiDjgBSnl/PPb\n/ICDQKyUsl4I8THwNm3DMB226cKAJHeAzfuz+HRvFldNCuTOa6L6/Pj1LfV8mr6Vg+eOoqExyj2Y\nFWOvIcorot+TfHJ2Oe9/nkZBaR0ujvbcND+Mq+ICCfAfYTN/HK3mVrZl72Znzh7Mmpn5QbO5Mfy6\nDse6e0LTNIrrS0ivahvGSa/MoqKxkgnGcVwVPIcIz7Ar/vmo5NO/1PntXG+nQj4JLABMwAPAZKBS\nSrlRCHEn8CDQQttUyPs7aiOlTOzmaQYsubeazPzfm0cpKKnjV7dPJjKkf6YxFtUVsyVrJyeKTwEQ\n4RnGirHXEDZiTJ8/V2llAx9+kc7x1BJ0wIJJgdw4PwwPl7aEaCt/HPk1Z3n7zIcU1J7Dy9GTO6Jv\n7tda5iazqU+Gzmzl/NoqdX47N+wKh3UnvaCKv7xznAAfF564ezoG+/4bqsirKWBz5g5Ol6UAMN4n\nmhvClhHs3vsx/6YWE9sO5bDtcC4trWbCg0dw+9WRjA64dHyyL/44NE3rt08eJrOJnTlfsi37c0ya\nidkjp7MqYjnOVjjzqCMq+fQvdX47p5J7B97dKfniRAEr54aycq7lY7lXKr0yi00Z28moygJgqv8k\nrg9dgp+LscfH0jSNY7KED79Io7y6CU83B25ZGM6MGP8OE7Alfxwt5lYqGispb6ygrLGc8oYKyhrb\n/pU3VlDdXEOAix+hI0YTdv6f0dm31wn/XF0Rbyd/SG5NPiMcPLg9eg3jfPp+uKw/qeTTv9T57ZxK\n7h1oaGrl/712mJr6Zp64ezqBvq79/pyapnGmPJVNmdvJqylAr9Mza+RUrh1ztcV3ueYV1/LerlRk\nXiX2djqWTR/F9bNG4+TQ+eQno9Gds0UVXSbvqqZqtA4mNenQ4eXkiZvBlXN1RbSYv1mlys3g2i7Z\nj2GUe7DF63yaNTO7c79mS9ZOWs2tTA+YzM0RK3Dp45uRBoJKPv1Lnd/OqeTeifjUEl5cn0h48Ah+\ndftk9AM0q0XTNOJLEtmSuZOi+mLs9fbMD5rF0tELcXfouP5NbUMLG/Zmsie+AE2DSeG+rF0cjp+X\nyxgN6jEAAAdbSURBVCXHrW6u4dz/b+/eYts+yziOf500seMktRPHbtKkLYO0zwaFsZaLatK6QosY\nd6uKGAeBAO0GhgQSN4CQOAh2g+gFEwgkGGKTChpojE3qDqxiq0DTxqZBN2CP1qbNaJpD06Zujm3i\nmIu/kzohpxI7/8b7faTIfzunR6/sn1+/9v95R/vpGx2gd6yfvtF+Ll4ZYmg8u2B4V0WqaIomaI41\nkYo10xxL0lzXTCrWRCrWRDKamF2Xzk3n6BnppSvbTVf2DF3Z7jkbTlRFqtjS2D4b9u9MbCMZ/d+d\nmQbGzvPwvx+hK9tNY00Dn7z5ELem37PaYQ2Nwqe8NL6LU7gv4Sd/eI1X/Dyf/Yix77a17fSYm87x\nUv+rHD39Jy5ODBGtruVDW/ayf+sd1G0I9s6cns7z/N97ePR4F6MTU7Q2x/nE/k62tG8ohHg/vaMD\n9I0Fl+NTczcHjxChJd5EsjYZhHdd02xwN8eaSUY3rupNxaGJS5y+/NZs2P9nuGfOmaNN0eScsD+V\nPcMfTz3J5PQkuzLv454dB2moLf+rpnJS+JSXxndxCvclDA1f4Vu/eBHI8/1799DUGF3zGianp/jr\nuRd56swxhq+OUL8hzoe37aMtfwu/Pf4v+kYHqG0YZeu2CFV1I/SPDTCRuzLnb1RFqkjXtdBWn6G1\nfhNt8eAyE0/T3tq8Zg+Oq7lJ3ho+Oxv2p7Pdc3qcQ/AZ83t2HGT3plvXpKZyU/iUl8Z3cQr3ZTz3\nag8PPe3stjT3HXxvKDXk83l6LmQ5eup5Xh/5G7nIVfJ5mL9SVB2pZlM8TetMiNdvojWeIRNvYUPV\nwuvuYT448vk858cHZ5dyqiPV3PWOA+t628L5FD7lpfFd3Ntys47rsff9m3nhn3284ud5+Y0Bdlu6\n7CccTeWm6e4f5uTZLCfPZnmzJ8vl0atAI1TfQbT9DHWpS9yUamV7S8fsbLylLhVqW4PrFYlEyMTT\nZOJp9rR9IOxyRN4WNHMvcm5wlG8/+BK56TzR2mpaNsZIJWK0JGK0JOpmj1OJGI11Ndcd/iPjk5zs\nyRbC/BKn+4aZnLq2Np1sqKWzI8n29gSdHQm2ZBpK1ipAM5/y0viWl8Z3cZq5r8Dmlnq+ePdO/nKi\nl8HsBBcuj9MzuHA7nNqaKlIbg9BvKQr9VOGJoDFew8DQeBDkPZd482yW3gtjs78fAToyDXR2JIIw\nb0+QSsRuyOZdIrL+KNzn2bUjza4d104qGpuYZDA7EYR94XIwOz57XBzYxaoiEaaLXhVFa6q5ZVsT\n2zuCWfm7Nieoi2r4RaQ8lC7LiMdq2BqrYeumhd/8G5uY4sLlIPBnngAuZCe4OHyFdDJGZ3uC7R1J\nOjL1VFetj26MIrL+KdxXKR7bQDwWbG4hInKj0FRSRKQCKdxFRCqQwl1EpAIp3EVEKpDCXUSkAinc\nRUQq0Io+Cmlmh4E9wDTwVXd/ueh7HwTuB6YAd/d7zexO4HfA6wQnY55w96+UungREVnYsuFuZnuB\nTne/3cxuBh4Ebi/6kZ8B+9y918weMbO7gHHgOXf/eFmqFhGRJa1kWWY/8BiAu78BJM2s+Iyd3e7e\nWzg+D6QKx2qSIiISkpWEeytBaM8YLNwGgLuPAJhZG3AAOFr41rvN7DEzO25mB0pUr4iIrMBKwn3+\nDDwCczfiNLMM8DjwJXcfAt4EvuPudwOfA35pZmp1ICKyRlYSuD0UzdSBzUDfzBUzaySYrX/T3Y8B\nuPs5gjdUcfcuM+sD2oHuJf5PJJ2unJ15bkQa3/LS+JaXxvf6rGTm/gzwMQAzuw3ocffiJueHgcPu\n/szMDWb2KTP7WuG4FcgQPEmIiMgaWNFOTGZ2P3AnkAPuA3YBlwiC/yLwAteWa44Avyl8JYEagiWa\np8tQv4iILOBG2mZPRERKRGeoiohUIIW7iEgFUriLiFSg0D97vlTfGlk99fkpDzPbSXDm9mF3/6mZ\ndQAPE0yYeoHPuPtkmDWuZwuM76+A3QQnUQL80N2fDK3AdSDUcF9B3xopDfX5KSEziwM/Bp4tuvl7\nwAPu/qiZ/QD4AvDzMOpb7xYZX4Cvu/vRBX5FFhD2ssxyfWukNNTnp7QmgI8SzNBn7AOeKBw/QdCK\nQ/4/C42vXKeww33JvjVSMurzU0LuPu3uV+bdXF+0DDMAtK1xWRVjkfEF+LKZHTOzI2bWvOaFrTNh\nh/uyfWtk1dTnZ20U3291Py69hwiWZfYD/wC+G3I9N7yww33JvjWyeu5+zt1n+/wQjG97uFVVpBEz\nixaO29GSQkm5+5/d/UTh6uPAzjDrWQ/CDvfl+tbIKqnPz5p5FjhUOD4EPBViLRXHzH5vZjcVru4j\n+PSXLCH09gPz+9a4+2uhFlRhCm9QH0F9fkrGzHYBPwK2AZMET5afBn4NRAm6n37e3XOhFbmOLTK+\nDwDfAEaBEYLxHVz0j0j44S4iIqUX9rKMiIiUgcJdRKQCKdxFRCqQwl1EpAIp3EVEKpDCXUSkAinc\nRUQqkMJdRKQC/RfO8Z0I0m6K6gAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fed4957aa50>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "df_val_loss[['org_val_loss', 'small_val_loss']].plot()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {
    "collapsed": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXIAAAEDCAYAAAAoWo9tAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fed49556090>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "df_val_loss[['org_val_loss', 'big_val_loss']].plot()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 2",
   "language": "python",
   "name": "python2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.13"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}