part2-IMDB-Classify.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Part2-进行IMDB影评的分类\n",
    "这个任务为二分类任务，分为positive和negative两类"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.5/dist-packages/h5py/__init__.py:36: FutureWarning: Conversion of the second argument of issubdtype from `float` to `np.floating` is deprecated. In future, it will be treated as `np.float64 == np.dtype(float).type`.\n",
      "  from ._conv import register_converters as _register_converters\n",
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "from keras.datasets import imdb\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Dense\n",
    "\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 导入数据\n",
    "# num_words=10000表示我们只需要top10000经常出现的word\n",
    "(train_data, train_labels), (test_data, test_labels) = imdb.load_data(num_words=10000)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(25000,) (25000,)\n",
      "(25000,) (25000,)\n"
     ]
    }
   ],
   "source": [
    "# 我们来看一下数据的shape,25000个样本的训练集和25000个样本的测试集\n",
    "print(train_data.shape, train_labels.shape)\n",
    "print(test_data.shape, test_labels.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "218 189\n"
     ]
    }
   ],
   "source": [
    "# 从这里看出，每个样本的长度是不一致的\n",
    "print(len(train_data[0]), len(train_data[1]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[1, 14, 22, 16, 43, 530, 973, 1622, 1385, 65]"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# 这些数字都是word的representation\n",
    "train_data[0][:10]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 在这里，我们也可以将index->word\n",
    "word_index = imdb.get_word_index()\n",
    "# 将word和index创建字典\n",
    "reverse_word_index = dict(\n",
    "[value, key] for (key, value) in word_index.items())\n",
    "# 这里我们为什么要从3开始呢，因为0：为填充， 1为序列的开始，2为unknown\n",
    "decode_review = ''.join([reverse_word_index.get(i - 3, '?') for i in train_data[0]])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### data process"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 这一步是为了将数据vector化，以便后面可以input到网络中\n",
    "def vectorize_sequence(sequences, dimensions=10000):\n",
    "    # 创建一个(n_samples, n_dim)的矩阵\n",
    "    results = np.zeros((len(sequences), dimensions))\n",
    "    # 这一句代码的意思是，因为是10000个word，一段影评中，一般可能全部词语都出现，\n",
    "    # 所以我们将出现的词语的index的置1，没有出现的为0\n",
    "    for i, sequence in enumerate(sequences):\n",
    "        results[i, sequence] = 1.\n",
    "    return results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "x_train = vectorize_sequence(train_data)\n",
    "x_test = vectorize_sequence(test_data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(25000, 10000)\n"
     ]
    }
   ],
   "source": [
    "print(x_train.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "## 这里要将label与vector/array化\n",
    "y_train = np.asarray(train_labels).astype('float32')\n",
    "y_test = np.asarray(test_labels).astype('float32')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(25000,)\n",
      "1.0\n"
     ]
    }
   ],
   "source": [
    "print(y_train.shape)\n",
    "print(y_train[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 按照7：3的比例分割training_set和val_set\n",
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "x_train, x_val, y_train, y_val = train_test_split(x_train,y_train, test_size=0.3, random_state=42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(17500, 10000)\n"
     ]
    }
   ],
   "source": [
    "print(x_train.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Buliding the Network"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = Sequential()\n",
    "# 因为输入层是10000个特征，所以input_shape为(10000, 16)\n",
    "model.add(Dense(units=16, activation='relu', input_shape=(10000, )))\n",
    "model.add(Dense(units=16, activation='relu'))\n",
    "model.add(Dense(units=1, activation='sigmoid'))\n",
    "\n",
    "# compile\n",
    "# 当然，你也可以选择其它的优化器，例如Adam/SGD..\n",
    "# 你也可以选择mse等其它的loss function， 但是我认为binary_crossentropy是二分类较好的loss function，因为最后输出的是概率\n",
    "model.compile(optimizer='rmsprop', loss='binary_crossentropy', metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 17500 samples, validate on 7500 samples\n",
      "Epoch 1/20\n",
      "17500/17500 [==============================] - 3s 147us/step - loss: 0.4988 - acc: 0.7954 - val_loss: 0.3681 - val_acc: 0.8704\n",
      "Epoch 2/20\n",
      "17500/17500 [==============================] - 2s 95us/step - loss: 0.2859 - acc: 0.9050 - val_loss: 0.3069 - val_acc: 0.8792\n",
      "Epoch 3/20\n",
      "17500/17500 [==============================] - 2s 95us/step - loss: 0.2107 - acc: 0.9306 - val_loss: 0.2827 - val_acc: 0.8796\n",
      "Epoch 4/20\n",
      "17500/17500 [==============================] - 2s 97us/step - loss: 0.1670 - acc: 0.9431 - val_loss: 0.2896 - val_acc: 0.8776\n",
      "Epoch 5/20\n",
      "17500/17500 [==============================] - 2s 100us/step - loss: 0.1403 - acc: 0.9535 - val_loss: 0.2892 - val_acc: 0.8820\n",
      "Epoch 6/20\n",
      "17500/17500 [==============================] - 2s 97us/step - loss: 0.1137 - acc: 0.9637 - val_loss: 0.3199 - val_acc: 0.8759\n",
      "Epoch 7/20\n",
      "17500/17500 [==============================] - 2s 98us/step - loss: 0.0957 - acc: 0.9697 - val_loss: 0.3332 - val_acc: 0.8756\n",
      "Epoch 8/20\n",
      "17500/17500 [==============================] - 2s 96us/step - loss: 0.0831 - acc: 0.9739 - val_loss: 0.3680 - val_acc: 0.8727\n",
      "Epoch 9/20\n",
      "17500/17500 [==============================] - 2s 96us/step - loss: 0.0686 - acc: 0.9790 - val_loss: 0.3892 - val_acc: 0.8715\n",
      "Epoch 10/20\n",
      "17500/17500 [==============================] - 2s 96us/step - loss: 0.0554 - acc: 0.9844 - val_loss: 0.4499 - val_acc: 0.8652\n",
      "Epoch 11/20\n",
      "17500/17500 [==============================] - 2s 97us/step - loss: 0.0470 - acc: 0.9866 - val_loss: 0.5008 - val_acc: 0.8568\n",
      "Epoch 12/20\n",
      "17500/17500 [==============================] - 2s 97us/step - loss: 0.0407 - acc: 0.9893 - val_loss: 0.4934 - val_acc: 0.8644\n",
      "Epoch 13/20\n",
      "17500/17500 [==============================] - 2s 97us/step - loss: 0.0322 - acc: 0.9926 - val_loss: 0.5100 - val_acc: 0.8665\n",
      "Epoch 14/20\n",
      "17500/17500 [==============================] - 2s 98us/step - loss: 0.0234 - acc: 0.9950 - val_loss: 0.5813 - val_acc: 0.8621\n",
      "Epoch 15/20\n",
      "17500/17500 [==============================] - 2s 97us/step - loss: 0.0219 - acc: 0.9947 - val_loss: 0.5834 - val_acc: 0.8624\n",
      "Epoch 16/20\n",
      "17500/17500 [==============================] - 2s 97us/step - loss: 0.0158 - acc: 0.9969 - val_loss: 0.6220 - val_acc: 0.8617\n",
      "Epoch 17/20\n",
      "17500/17500 [==============================] - 2s 97us/step - loss: 0.0138 - acc: 0.9971 - val_loss: 0.6617 - val_acc: 0.8609\n",
      "Epoch 18/20\n",
      "17500/17500 [==============================] - 2s 95us/step - loss: 0.0123 - acc: 0.9978 - val_loss: 0.6985 - val_acc: 0.8613\n",
      "Epoch 19/20\n",
      "17500/17500 [==============================] - 2s 98us/step - loss: 0.0064 - acc: 0.9993 - val_loss: 0.7936 - val_acc: 0.8544\n",
      "Epoch 20/20\n",
      "17500/17500 [==============================] - 2s 94us/step - loss: 0.0088 - acc: 0.9985 - val_loss: 0.7750 - val_acc: 0.8577\n"
     ]
    }
   ],
   "source": [
    "history = model.fit(x=x_train, y=y_train, batch_size=512, epochs=20, validation_data=(x_val, y_val))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "dict_keys(['val_acc', 'loss', 'val_loss', 'acc'])\n"
     ]
    }
   ],
   "source": [
    "# 查看模型训练返回的结果\n",
    "history_dict = history.history\n",
    "print(history_dict.keys())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAmQAAAHjCAYAAACNTANBAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMS4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvNQv5yAAAIABJREFUeJzs3Xd8leX9//HXdVb2DkmAAGEvAQUUq2JFWVoVd121tY4ObWv9/mxtbfttq9126dc62rpHtbUijgqKUvcABwoCsgkQAoQMMs+4fn/cJychBAiQk/skeT/b+3HPc87npCW8ua7rvm5jrUVERERE3ONxuwARERGR3k6BTERERMRlCmQiIiIiLlMgExEREXGZApmIiIiIyxTIRERERFymQCYiIiLiMgUyEREREZcpkImIiIi4zOd2AQcrPz/flpSUuF2GiIiIyAEtWbJkh7W2z4Gu63aBrKSkhMWLF7tdhoiIiMgBGWM2dOQ6dVmKiIiIuEyBTERERMRlCmQiIiIiLlMgExEREXGZApmIiIiIyxTIRERERFymQCYiIiLiMgUyEREREZcpkImIiIi4TIFMRERExGUKZCIiIiIuUyATERERcZkCmYiIiIjLFMhEREREXBa3QGaMudcYU26M+WQf540x5jZjzGpjzFJjzMR41SIiIiKSyOLZQnY/MHs/508FhkeXq4E741iLiIiISMLyxeuNrbWvGmNK9nPJHOBBa60F3jbGZBtj+lprt8arJhEREUk8ERshGAkSDAexWLzGi8d48Hq8se2eLm6BrAP6A5ta7ZdGj+0VyIwxV+O0ojFw4MAuKU5ERKSjwpEwjeFGmsJNhGwIay0WG1sDLcdaHd9rjcX5r93renCCS+vrIjZChIiz3sdisYQj4T2us9YStuG9r7PhPc613g5GgoQioVhwCkbaLNFjsWv2czwUCe3xHmEbPuDP2GuccOb1RMOaaQlrbY97jAefx7fndZ5W54xz7pwR5zC7ZH+deV3HzUDWYdbae4B7ACZPnmxdLkdERA4gFAnRGG6kIdTgrMMNNIYaaQw3Uh+qj4UXi8VgMMYQ+4/Zew3sdZ3zX4PHePZ9XXRtsTSGG2M1tF6awk00hBtoCjftcU3b43td12oJRUJu/ri7lM/jw+/xtyxeZ93e8XR/Ov6kNue9/j2u83l8exwDJ2g2B8FwJNyybcN777e5rnXQDEVCe7yueTsYCRKyTihMFG4Gss3AgFb7xdFjIiISR8FwkLpQHbXBWmqDtbHtumAddaE6GkINsdDUHKjahqvYdnvnQw2EbPcLKD7jI8mXRJI3iYA3QLI3mYA3QJLXOZYRyCDPm7fn8VbXN1/nM76WcNhO0Gw+3hwk93ldNHA2H/Pg2fuY8TgLHjweZ22MibUENS/Nxwwt52LHWl1vMHg9Ldc1f4bXeGPhqfm7SedyM5DNA641xvwDmAJUafyYiMieIjYSa1VqHZqat5sDVfN22/22Yas2WEsw0vFWAYMh2ZdMsjeZJF+Ss44GkWRvMhmpGSR5k9o9n+JNafc1Sd4kUnwp+L3+WOtVc5dc83aEyH677g7U7RfrKoy+h8HEwlPz0jpEBbwBfJ5u0WkkPVTc/t9njHkMOAnIN8aUAv8L+AGstXcBzwOnAauBOuDyeNUiIhIPwUiQhlBDbKkP17fsh51Wpnb3oy1M7e6HW71fdL+jfMZHqj+VNH8aqb7o2p9KXnJebDvVn0qaL22v/davS/G1BCm/x68WEZEuEM+7LC86wHkLXBOvzxeR3qcx3Miuhl3Uheqc8T6hhj3G/cTWB3t8H+c7MhC5LZ/HR4o3hRRfitPyFF1SvClkpGbEWqOSfckt10T30/xOkGoOUG3DV8AbiMNPVUS6gtpnRSQhRWyEmqYaKhoqqGysdNYNlexq3MWuhujSZrs+VH/Qn+M13pZurDZdWkneJDIDmfTx9mn3fMAbIMWXsldwSvGlxLZb7yf5kmKDlkVEWlMgE5Eu0dx61TpI7S9oVTVW7bMFKsWXQk5SDtnJ2eQk5zA4azA5yTmxY6m+1NjA62RfcmyAdnvBS+OGRCQR6DeRiBy0+lA9lQ2VVDZWxsLTrgZn3d6x/bVeGQzZSU6wyk7KZnDWYI5KPoqcpBwnZEWDVuvAleJL6eJvLCISXwpkIr2YtZbaYC2VjZWx4FTZWBkLW7GlzX5juHGf75kRyHACVlIO+Sn5DMseRnZydix05Sblxlq2cpJyyAxk4vV4u/Bbi4gkHgUykW6uKdxEdVM1NU011DTV7HO7vXPVTdX7nNDSYMhKyiI7yQlTfdP6MjpvdGw/Oyk7FrSal6ykLHUBiogcAv3mFEkA4UiYXY272Fm/k4qGCqqaqvYIUq3DVNuQtb/WKgC/x09mIJOMQAaZgUwyA5n0T+9PRiCDjEAGOUk5ZCVlxboMm5eMQIZarkREuogCmUicBMNBdjbsZGfDTirqK5zt+p17rSsaKtjVsCv2vLu2fMYXC0/NoaootSi23fZc2+NJ3iTNIyUikuAUyEQOQkOooSVQRUNVRUNFu0Gruqm63fdI8aWQm5xLXkoeAzIGcGTBkeQl55GXkkdech65ybmxFqqMQAYpvhQFKhGRHk6BTKSNXQ272FC9gXVV61hfvZ71VevZUL2BsroyaoO17b4m3Z8eC1TDsodxTNExsf2261R/ahd/IxERSXQKZNIrNYWb2Fi90Qle1etYX7XeCV/V66lqrIpd5/P4GJAxgJLMEo7td2y7ASs3JZckb5KL30ZERLo7BTLpsay1lNeVs6F6A+ur1+/R4rWldgsRG4ld2yelD4MyBzFj0AxKMksYnDWYkswS+qX3012DIiISd/qbRrq9umAd66udbsX1VetjLV4bqjdQF6qLXZfiS2FQ5iCOyD+CLwz5AiVZJQzOHMygzEGkB9Jd/AYiItLbKZBJt2CtZVvdNtZVrWtZqp11eV157DqDoV96P0oyS5hYOJGSzBIGZQ5icNZgClIL8BiPi99CRESkfQpkklAaw42xAfWtl/XV6/d49E66P50hWUM4tu+xlGSWUJLlBK+BGQNJ9iW7+A1EREQOngKZdDlrLRUNFbGg1Tp4bd69eY/5uPql9WNw1mAmFU5icNbg2JKXnKepIEREpMdQIJO4CUVClNaU7tG92Ly0nqMryZtESWYJR+QfwRlDz4iFrkGZg/QQaRER6RUUyOSwWWvZUruFFRUrWFmxklW7VrGuah0bazbu8ZzE/JR8BmcNZnbJ7FjoKskqoW9aX43tEhGRXk2BTA5KY7iR1ZWrWVWxyglgu1ayqmIVNcEawBlUPyhzEEOyhjBtwLQ9gldmINPl6kVERBKTApnsU0VDBSsqVjjha5fT+rWuah1hGwacaSRG5IzgtCGnMSJnBKNyRzEse5hmohcRETlICmRCOBJmY81GVlasZOWulbEQVl7fMp1EYWohI3NHMm3ANEbmjmRU7igGZAxQV6OIiEgnUCDrZeqCdazatSoWvlZWrOSzys9iU0r4jI8h2UOY0ndKLHiNzBlJdnK2y5WLiIj0XApkPVzERni37F2eWfMMH23/iI3VG2PTSmQGMhmZO5Jzh58bC19DsoYQ8AZcrlpERKR3USDrobbs3sLTq59m7uq5bKndQoY/gyl9p3D6kNMZmeOEr6K0Is3lJSIikgAUyHqQhlADCzcuZO7qubyz9R0ApvSdwncmfoeTB56sGexFREQSlAJZN2etZdnOZcxdPZfn1z5PTbCG/un9+caR32DO0Dn0S+/ndokiIiJyAApk3VRFQwXPrnmWp1Y/xerK1SR5k5gxaAZnDTuLo4uO1t2PIiIi3YgCWTcSioR4c8ubPPXZUyzatIiQDTEufxw/PvbHnDr4VDICGW6XKCIiIodAgawbWFe1jrmr5/LMmmfYXr+d3ORcLhl9CWcNO4thOcPcLk9EREQOkwJZgqoN1jJ//Xzmrp7LB+Uf4DVephZP5exhZzO1eCp+j9/tEkVERKSTKJAlEGst75e/z1OfPcWCDQuoD9UzOGsw10+6njOGnkF+Sr7bJYqIiEgcKJAlgG2125i3Zh5zV89lY81G0vxpnDb4NM4adhYT+kzQXGEiIiI9nAKZiz7d+Sm3fXAbb255k4iNcHTR0XxtwteYPnC6HtAtIiLSiyiQuaQ+VM91r1xHQ7iBK8ddyVlDz2JA5gC3yxIREREXKJC55J6l97Cldgv3zbqPyUWT3S5HREREXKTZQ12wtnIt9y+7nzOHnqkwJiIiIgpkXc1ayy3v3EKKL4XrJ13vdjkiIiKSABTIuthz657jvbL3uG7ideSl5LldjoiIiCQABbIuVN1Uze/e+x3j8sdx7vBz3S5HREREEoQG9Xeh29+/ncrGSu6cfidej9ftckRERCRBqIWsiyzbsYzHVz7OhSMvZEzeGLfLERERkQSiFrIuEI6Eufntm8lLyePao651u5xuyVqLra8nXF1NuKqaSHUV4ZoabFMQX14u3vx8fPn5eNLT9WQDERHpdhTIusA/V/2TZTuX8ZupvyEjkAFApKGBYGkp4cpKbCQCFrARsNbZj1jAQiSCtbb9fWtbvaZlP/Z+kQgQfT9rMV4fJikJkxTAEwg424EkTCCAJ6l5P7r2O8fw+Tot4NhIhEhtbUugqq4mXF1NJBqynP0qItU1LdvNx2tqIBg84GeYpCR8eXl4++Tjy8vHl5eHr08+3rw8fPl98OXn4cvPx5uXjyctVeFNREQSggJZnNhwmFB5OdtXf8KSZ37HdxuLmPDJItaXPkywtJTQ9u1ul9gxHk8spHkCgZbAlpSECfjxBJJaBbkAnkAS+LxEauuc0BULWtVEamqiIXEfvF68GRl4sjLxZmbhzcjA36+fs52ZiTcrE09mprOdmYknMwvj8xLauZPwzp2Etu+Ibu8gtH0Hwc2bqf/oI8IVFU5YbcMkJ+PLz98zwOXn48vPc1rc8vJjAc6TqkdZiYhI/CiQHYZwdTVNmzYRLN1MsHQTTaWlBDeVEty0ieCWLdhoi85XAUw9dX3DBIoHkDZ1KoEBxfiLi/Hl5YHHA8YDBozHA8aA8WA8xtn2eAADHuO06MSuaWe/eTv6mth7GIMNhbFNTdimRmxjI7apiUhjdL+pCdvYSKSxEdsUjJ5vJNLUhG1sip3f41j0PcI11XteEwziSU/Hm5GBNy+XwODBToDKysSb0TpYZeHNag5XmXjS0uLSYmVDIcK7dhGKhbYdewS40I7tBDdspH7J+4QrK9sPb6mpTigzYGj5mbYsh3DcGGjnnDEGk5yMJyUFT2oKJiUFT4rz+e0eS03Bk5KCSYluN1+XkoIJBDr95ykiIp1PgWw/bFMTwS1baNpUSnBzaUv42rSJps2biVRV7XG9JyuLQHExSaNGkTFjOpszw9y6+UFOmnIhV07/vv5ydInx+fD16YOvTx8Ytf9rbShEqKKC8I4dewa4HTuI1DdEw5p1uo2tjXY125Yldu4wjkciRBobCe+uIVReTqS+nkhdHZH6emx9/cF9eZ8vGuKckGZSo0EueswkJ2F8fozPh/H5wOd19r1ejN8HXm+r816nC9vrO+A5443ut1o8aWl48/Px6M+BiMheFMjaqFu8mO1/vo2m0lJCZWV7tJYYvx9///74Bwwgc8J4AsUD8BcXx1q7vJmZsWuD4SBXPXMejXkD+NLMGzA+/SXUHRifD39BAf6CArdLaZeNRLANDS0hra4eW++EtUh9vdNVXF+Hbd6vaw5z0WN1zvFwVRXBsq3Y+gZsOIwNhSAUwjYv4XCHxuwdCk9WVrRruE0XcX4fZ7+Pc9ybk+MEOxGRXkCBrC2PFxsOk3bM0fj7F+MfMIBAsRPCfAUFTpdgBzyw/AHWVq3ljlPuINmXHOeipbcwHk9L92lefJ/0YKOtdXuFtVAYQsFYkNvzfBgbCkKrc83nw7t3Oy2P23cQ2uEs9R9/7LQ+1tXtXYDHgzcvtyWoNS/Nga3VMU9Ghm7QOEjWWggGnWEK0aV5CIMJJDnDCTIy1LIv0kUUyNpInXgUJY8+cljvsXn3Zu7+6G5OGXgKJxaf2EmViXQtY4zTLen1QlJSXD8rUlsbHc/XHNi2E9qxY48A1/jZZ4R27IBQaO9aAwEnpPVxWtr8ffs6rdn9++Hv359A//54srK6XWgL764lVLaV4NYywpW7WsZ5NkTHczY2xsZzRpqi2w0NLdvNQasp+prG5jGgznZ74yXbMikpLTfStLrhJra9jxtuvFmZeJL1j1GRjlIgi4Nfv/NrjDHceMyNbpci0i140tIIpKURGDhwv9fZSIRwVZUT1HbsILQjGuJ2bI+Ft+DGjdS9/TaR2tq9PsMJaa2WYies+fv3x5OZ2aWBLdLQQKisjGBZGcGtZbHgFSzbSmirczxSU7P/N/F4nBtAWt397Gm+CzopCU9KMiYrq2W6m6TkVtvRaW/angv4sY2N7U9JU1VNcMsWGqLbbX/GbZlAoOVmntaBrjnEpWc4N7O0GlOJbWdcZfPx2BRBbcdg7uP61sfBuTHK4wWvx+nt8Hhbjnk8GK9zg1XzGq/H+QdJ22PR1xqPcd4jdqzVOW+rtdfrHPc2f0507fPFjrd73utt9z068v/T5hbu5nGpFpz96DHbasxqu/utplrCWvD68OXlOmNNJS70k+1kr2x8hUWli7h+0vUUpRW5XY5Ij2I8Hnw5OfhyckgaPnyf11lriVRXE9y8mabNmwlu3kxw8xZnXVpK3bvv7h3Y0tPbBLaW1rWDDWw2GCS4rbzdkNW8Hd61a6/XeXNz8RcV4R84kNRjjsHftwhfUV/8fYvw5ubiSU7eM3S5/JejDYUI19Q4wa31pM3N2zV7Brrw9h00rVkbe01HWugOaK87mA2mvePgBJPmUBIOH/5nu6V1iGsnTHXKz7U9xuDNz8NfUIivoABfYQG+6JhbX2H0WEEB3uzsbtcanQiMjdf/cHEyefJku3jxYrfLaFddsI6znj6LNH8aT5zxBH6P3+2SRKQd1loiVVV7h7XmpbR0r3FtscBWXIy/fz8C/fvjzc1zWuhiwauM0NatTtdqm9+tnsxM/EVF+PoW4Y+GLF/RntueOHcNJxIbiTg/Y2ujgclEV+1PBdPuscOtwVonmLUKajYa1mLBrfW5cMSZjDscdlqQIuHYMRsOOxN2Nx9rdx39rNbr8D7OhyPYyH7WoTA2Eo6Oa26e/oj296NTKcWmTjLN0yPtvR+bPql5qqXo620wSKh8O8HybYTKywltKydUXt7uPyxMIBALZ77CaGBrFeL80XOdNb+jtdbpjq+tdW5iqq1tWVrv73GujkhdLZlnnknmjBmdUse+GGOWWGsnH+g6tZB1onuW3sPW2q3cP/t+hTGRBGaMwZudTUp2Niljx+51fq/AVtoqrG3cSO1bb2FbBTaTmuq0bBUVkXTi1FYhy1n7i4rwpKV15VdMeMbjwZue7m4NxkC0lVHtOYcm0tREqHw7oeagVl5OcNs259i2bTR+uoLd/311jz8vzTwZGU4LW2EBvj4trWwmKbCPMFW3z9DV0RZP4/fjSUtzltRUIjW7O/tHcsgUyDrJmso1PLDsAeYMncOkwklulyMih6EjgS1cWUm4osK5y7OLx5+JJApPIECguD+B4v77vMZa69y4s23bXoEtVF5OsHwbjevedZ5g0+amHRMI7BGgPGlpeDMz8fftG9tve96Tlrrn9a3OJ/JdwwpkncBayy1v30KqP5XrJ1/vdjkiEmfGmNhYNhHZP2MM3vR0vOnpJA0dus/rbCRCeOdO52kvzQHK33t6mxTIOsGza59l8bbF/ORzPyE3OdftckRERLod4/E4T1TppTo2y6nsU1VjFbcuvpXx+eM5d/i5bpcjIiIi3ZBayA7T7R/cTmVjJXdNvwuPUb4VERGRg6cEcRg+2fEJT6x8gotGXcTovNFulyMiIiLdlALZIQpHwtz89s3kp+Rz7ZHXul2OiIiIdGMKZIfoiVVPsHzncm44+gbSA+7OpSMiIiLdmwLZIdhRv4Pb3r+NY/sey+yS2W6XIyIiIt2cAtkhuHXxrTSGG7lpyk2aDFJEREQOmwLZQXpn6zs8t/Y5vnrEVynJKnG7HBEREekBFMgOQlO4iVvevoXi9GKuHHel2+WIiIhID6F5yA7CA8seYH31ev5yyl9I9iW7XY6IiIj0EGoh66DSmlLuXno3MwbNYGrxVLfLERERkR5EgawDrLX86t1f4TEevnf099wuR0RERHoYBbIOeHnTy7xa+irXHHkNRWlFbpcjIiIiPYwC2QHUBev49bu/Zlj2MC4efbHb5YiIiEgPpEH9B3DX0rsoqy3jgdkP4Pf43S5HREREeiC1kO3H6l2reWjZQ5w17CwmFk50uxwRERHpoeIayIwxs40xK40xq40xN7ZzfqAx5hVjzAfGmKXGmNPiWc/BsNZyyzu3kBZI4/pJ17tdjoiIiPRgcQtkxhgvcAdwKjAGuMgYM6bNZT8CnrDWHgVcCPwlXvUcrGfWPsOSbUu4buJ15CTnuF2OiIiI9GDxbCE7BlhtrV1rrW0C/gHMaXONBTKj21nAljjW02FVjVX8fvHvmdBnAucMP8ftckRERKSHi+eg/v7Aplb7pcCUNtf8FFhgjPkWkAZMb++NjDFXA1cDDBw4sNMLbeu292+jsrGSe2bcg8domJ2IiIjEl9tp4yLgfmttMXAa8JAxeycga+091trJ1trJffr0iWtBH2//mH+u+icXj7qYkbkj4/pZIiIi0kpTLdTvcrsKV8SzhWwzMKDVfnH0WGtXALMBrLVvGWOSgXygPI517Zff62dq8VSuOfIat0oQERHpfYIN8PdZsO0T6D8Rhp4Cw06B/pPB2/Nn6YpnC9l7wHBjzGBjTABn0P68NtdsBE4BMMaMBpKB7XGs6YBG5Y7ijlPuID2Q7mYZIiIivcvCn8O2j+HoK8B44bVb4d5Z8Nsh8PilsPg+qNzodpVxE7fIaa0NGWOuBeYDXuBea+0yY8zPgcXW2nnA/wB/NcZ8F2eA/1estTZeNYmIiEgCWvMKvH0HHHM1nPY751j9Llj7X1j9Eqx5GT59xjmeP6Kl9WzQ8RBIda/uTmS6W/6ZPHmyXbx4sdtliIiISGeoq4A7j4OkTLh6UfsBy1rYvhLWLITVC2HDGxBqAG8SDPpcS0ArGAPGdPU32C9jzBJr7eQDXqdAJiIiIq6wFv75ZVjxPFy1EPpO6NjrgvWw4U2n5Wz1S7B9hXM8o280nJ0MQ6ZBam78au+gjgaynj9KTkRERBLTR4/B8qdh+s86HsYA/ClOi9iwU2DWL6CqNBrOFsKKZ+HDhwHT6uaA6dB/UkLfHKAWMhEREel6FevgrhOg75Hw5Xng8XbO+0bCsPn96NizhbB5CdgIJGXBkM87IW7oKZA94MDv1QnUQiYiIiKJKRyCp77m3E159l2dF8bAea8BRzvLtB84Y9TW/ddpPVvzMnwanfAhfwSc+D0Yf37nffZhUCATERGRrvX6H2HTO3Du3+PfUpWaC2PPdpbmmwOaW8+8/vh+9kFQIBMREZGuU7oEFv0Kxp0P487r2s82BgpGOctx13btZx+A249OEhERkd6icTf8+yrI7Aen3ep2NQlFLWQiIiLSNRbcBBVr4SvPQkq229UkFLWQiYiISPyteB6W3A/HfwdKTnC7moSjQCYiIiLxVbMN5l0LReNh2k1uV5OQFMhEREQkfqyFp6+Bplo492/gC7hdUULSGDIRERGJn/f+BqtfdAbx9xnpdjUJSy1kIiIiEh/bV8KCH8GwGXD0lW5Xk9AUyERERKTzhZrgySshkAZz7nDmAJN9UpeliIiIdL5Fv4SypXDho5BR6HY1CU8tZCIiItK51r8Or/8JJn4ZRn3B7Wq6BQUyERER6Tz1lfDU1yF3CMz+ldvVdBvqshQREZHO8/wNUL0FrnjRGT8mHaIWMhEREekcH/8LPn4CTroRiie5XU23okAmIiIih69yEzx7PQyYAidc73Y13Y4CmYiIiByeSNgZN2bDcPbd4NWIqIOln5iIiIgcnjdvhw2vw5y/QO5gt6vpltRCJiIiIodu60fw8i0wZg4cebHb1XRbCmQiIiJyaIL18ORVkJYPp/9Js/EfBnVZioiIyKF58SewYyV86SlIzXW7mm5NLWQiIiJy8D57Ed69B469Boae7HY13Z4CmYiIiByc2h3w9DVQMAZO+Ynb1fQI6rIUERGRjrMW5n0b6nfBpf8Gf7LbFfUICmQiIiLSce8/CCufg1m/hKIj3K6mx1CXpYiIiHTMzjXwwo0w+PMw5RtuV9OjKJCJiIjIgYWD8O+rwBuAs+4EjyJEZ1KXpYiIiBzYq7+DzUvg/Acgq7/b1fQ4irciIiKyf5vedQLZhIth7FluV9MjKZCJiIjIvjVUO12VWQPg1N+4XU2PpS5LERER2VskDB89Bi//AnaXweX/geRMt6vqsRTIREREpIW1sPol57FI5cuh/2Q4/z4YeKzblfVoCmQiIiLi2PKBE8TWvQo5g50B/GPm6KHhXUCBTEREpLfbtQFevhk+/iek5sGpv4VJl4Mv4HZlvYYCmYiISG9VVwGv/d55SLjxwNT/geO/A8lZblfW6yiQiYiI9DbBBieEvXarcxflkZfAtB9qfjEXKZCJiIj0FpEIfPIvWHgzVG2EYTNgxs+gcKzblfV6CmQiIiK9wdpFsODHULYU+k6AObfDkJNcLkqaKZCJiIj0ZNuWOXdOrn4JsgbCOX+FI87TsygTjAKZiIhIT1S1GV75JXz4iDOh68xb4OirwJ/sdmXSDgUyERGRnqShCl7/E7z9F7AR+Nw1zt2TqbluVyb7oUAmIiLSE4SaYMl98N/fQN1OGHcBnPwjyBnkdmXSAQpkIiIi3Zm1sHwuvPQz2LUOBp8IM34O/Y5yuzI5CApkIiIi3dWGN507JzcvhoIxcMm/YNh0PeqoG1IgExER6W52rnGC2MrnIKMvzLkDJlwEHq/blckhUiATERHpLppq4bU/wJu3gTcAJ/8Yjv0mBFLdrkwOkwKZiIhIomseJzb/R1BdCuO/CNN/Bpl93a5MOokCmYiISCIrXwH/uQHWvQqF4+Dcv8Kg49yuSjohmQa2AAAgAElEQVSZApmIiEgiaqiCRb+Bd++GQBqcditMuhy8+qu7J9L/qiIiIokkEoGljzuPO6rdDhMvg1N+Amn5blcmcaRAJiIikii2fAjP3wCl70L/yXDx49B/ottVSRdQIBMREXFbXQW8fDMsvg9S86LTWFysB4D3IgpkIiIibomEYcn9ThhrqIYpX4eTboSUbLcrky6mQCYiIuKGje/A8/8PypbCoBPgtN9C4Vi3qxKXKJCJiIh0pZpt8NL/wkePQUY/OO9eGHuOHnfUyymQiYiIdIVwEN69B175FYQa4ITvwtT/B0npblcmCUCBTEREJN7WLoL/fB+2r3Ae/j37N5A/zO2qJIEokImIiMRL5SZYcBMsfxqyB8GFj8HIU9U9KXtRIBMREelswQZ463Z49feAhWk3wXHfAn+K25VJglIgExER6UwrX4AXvg+71sPoM2DmLyBnkNtVSYJTIBMREekMOz6D+T+EzxZA/gj40lMw9GS3q5JuQoFMRETkcNRXwqu/g3fuAl8KzLjZmeDVF3C7MulGFMhEREQORSQM7z/ozLJfVwETvwQn/xjSC9yuTLohBTIREZGDte41eOEHsO1jGPg5mP1r6Hek21VJN6ZAJiIi0lG7NsCLP3amscgaAOfdB2PP1jQWctgUyERERA6kcTe8/gd48//A49U0FtLpPPF8c2PMbGPMSmPMamPMjfu45gJjzHJjzDJjzKPxrEdEROSgRCLw4WNw+yR47fcwZg5cuxg+/z2FMelUcWshM8Z4gTuAGUAp8J4xZp61dnmra4YDPwCOt9buMsZoJKSIiCSGTe8584ltXgL9JsIXH4IBx7hdlfRQ8eyyPAZYba1dC2CM+QcwB1je6pqrgDustbsArLXlcaxHRETkwKq3wEs/haWPQ3oRnHUXjP8ieOLaqSS9XDwDWX9gU6v9UmBKm2tGABhj3gC8wE+ttS/EsSYREZH2BeudMWKv/8GZ0mLq/8AJ10NSutuVSS/g9qB+HzAcOAkoBl41xoyz1la2vsgYczVwNcDAgQO7ukYREenJrIXlc2HBT6BqI4w+E2beDDklblcmvUg8A9lmYECr/eLosdZKgXestUFgnTFmFU5Ae6/1Rdbae4B7ACZPnmzjVrGIiPQuW5fCCzfChjeg8Ag461kYPNXtqqQXimcgew8YbowZjBPELgQubnPNXOAi4D5jTD5OF+baONYkIiICu7c7M+y//yCk5sLpf4KJlzlTWoi4IG6BzFobMsZcC8zHGR92r7V2mTHm58Bia+286LmZxpjlQBi4wVq7M141iYhILxdqgnfvhv/+FoJ1cOw3nSksUrLdrkx6OWNt9+oBnDx5sl28eLHbZYiISHdiLayaD/N/CBVrYPhMmPVLyB/udmXSwxljllhrJx/oOrcH9YuIiHRcOAThRgg1QqghujQeeL3yP7BmIeSPgEuehOHT3f4mIntQIBMRka5XuxNeuxVqd3Q8VIUawIYP7fOSs5wHgB99JXj9nftdRDqBApmIiHSthip4+BzYtgyyisGXDL6klnVy1p77e6zbO3agdTIkZYAv4PY3F9knBTIREek6TXXw6IWw7RO48FEYMcvtikQSggKZiIh0jVATPHEZbHwLzvu7wphIKwpkIiISf5Ew/PsqWP0inPFnOOJctysSSSh6UqqIiMSXtfDMd5zHE824GSZ9xe2KRBKOApmIiMSPtTD/JvjgITjxBjj+225XJJKQFMhERCR+/vsbePsOOOZrMO0mt6sRSVgKZCIiEh9v/QUW/QqOvMSZA8wYtysSSVgKZCIi0vnefwjm/wBGnwln3AYe/XUjsj/6EyIiIp1r2VPwzLdh6Mlw7t/Aqxv6RQ5EgUxERDrPZy/Ck1dB8THwxYed2fJF5IAUyEREpHNseBMe/xIUjIZLnoBAmtsViXQbCmQiInL4tnwAj1wA2QPgS085z6MUkQ7rUCAzxlxjjMlutZ9jjPlm/MoSEZFuo3wFPHQOpOTAl+ZCWr7bFYl0Ox1tIbvKWlvZvGOt3QVcFZ+SRESk29i1Hh46C7x+uGwuZPV3uyKRbqmjgcxrTMsEMsYYLxCIT0kiItItVG+FB+dAsN7ppswb6nZFIt1WR+9FfgF43Bhzd3T/a9FjIiLSG9XudFrGanfAZfOgcKzbFYl0ax0NZN8Hrga+Ed1/EfhbXCoSEZHE1lANj5wLFevg0ieheJLbFYl0ex0NZCnAX621d0GsyzIJqItXYSIikoCC9fDYhVD2MXzxERg81e2KRHqEjo4hW4gTypqlAC91fjkiIpKwQk3wxGXOfGNn3w0jZ7tdkUiP0dFAlmyt3d28E91OjU9JIiKScCJheOpq+GwBnP5HGHee2xWJ9CgdDWS1xpiJzTvGmElAfXxKEhGRhGItPHud84zKGTfD5Mvdrkikx+noGLLrgH8aY7YABigCLoxbVSIikhishQU/gvcfhBNvgOO/7XZFIj1ShwKZtfY9Y8woYGT00Mr4lSQiIgnjv7+Ft/4PjvkaTLvJ7WpEeqwOP8vSWhsElgGFwF1AabyKEhGRBPD2nbDolzDhYpj9a2iZH1xEOllHn2V5rDHmNmAD8DTwKjAqnoWJiIiLPngYXrgRRp8BZ94Ong7/+11EDsF+/4QZY35pjPkM+AWwFDgK2G6tfSD6PEsREelpls2Fed+CoSfDuX8Hb0eHG4vIoTrQn7IrgVXAncAz1tpGY4yNf1kiItLlIhFY+Tw8eSUUHwNffBh8SW5XJdIrHCiQ9QVmABcBfzLGvAKkGGN81tpQ3KsTEZH4iURg2yew4Q1Y/7qzrt8FRePg4schkOZ2hSK9xn4DmbU2jPMQ8ReMMUnA6Tiz9G82xiy01l7cBTWKiEhniISdRx41h68Nb0JDpXMupwRGfgFKjodRp0NypqulivQ2+w1kxpjPAW9bRyPwJPCkMSYTOKsrChQRkUMUDkHZ0lYB7C1orHLO5Q5xBuyXTHVCWFaxu7WK9HIH6rK8DLjDGLOKaEuZtbbMWlsNPBj36kREElXlJvjwEUjKhMx+TqDJ7AfpheDxulNTOARbP9wzgDXVOOfyhsERZ8OgE5wAltnPnRpFpF0H6rL8BkB0UthTgfuNMVnAKzgB7Y1ot2aP8eaaHfziuU958KvHkJeuwawi0o5V8+GprznjrdoyXsgocgJPZj/I7B9d+rWsM4rA6z/8OsJB2PKBE8DWvw6b3oGm6GOH80fA+PNh0PFQcoLzmSKSsDo6U/8KYAXwR2NMCjANOB/4AzA5fuV1vcxkP8u2VLPw03IuOHqA2+WISCIJh+CVW+D1P0LhOLhyIaTkQPVmqN7Sah3dLv8UPnsJgrVt3sg4LWmtQ1tW6+DWDzL67n2HY6gJtrzfKoC92/LefUbBhAudADboeMgo7JIfiYh0jg4FMmPMUKA0Oo5sCjAM+LG1tjKexblhbL9M+menMH9ZmQKZiLSo3gr/+ipsfBMmfcWZud6f4pxLzXXuTGyPtdBQtWdQa73euRrWvQqN1Xu/Nq1PS2BrqnUCWKjeOVcwFo66pCWApfeJy9cWka7R0dn+ngQmG2OGAffgzNb/KHBavApzizGGmWMLeeSdjexuDJGepAkRRXq9Na84c3MF6+Dse2DCFzv+WmMgJdtZCsfs+7qGaqjZuncrW/UWqNzovM+kLzvdjwOPg7S8w/9eIpIwOpo2ItbakDHmbOB2a+3txpgP4lmYm2aOKeK+N9bz6qrtnDaur9vliIhbImF49Xew6NfQZySc/wAUxOmpccmZztJnZHzeX0QSWkcfThY0xlwEfBl4NnqsE0akJqajS3LISfUzf1mZ26WIiFt2b4eHz4FFv4LxX4SrXo5fGBORXq+jLWSXA18HfmGtXWeMGQw8FL+y3OXzepg+upAXlpXRFIoQ8OmhuiK9yoY34Z+XO5OmnnEbTLzM6TIUEYmTDiUNa+1ya+23rbWPGWNygAxr7W/iXJurZo4toqYhxNtrd7pdioh0lUjEuYPy/tOdxwZd+ZIzbkthTETirEOBzBizyBiTaYzJBd4H/mqM+UN8S3PX1OH5pAa86rYU6S3qKuCxC+Glnzoz2F+9aN93ToqIdLKO9sVlRWfnPwd40Fo7BZgev7Lcl+z38vkRfXhx+TYiEet2OSIST6WL4e4TYc3LcOrv4Pz79SxHEelSHQ1kPmNMX+ACWgb193gzxxZSXtPIh6U9bro1EQFnjrC374R7ZwMGrpgPU65WF6WIdLmOBrKfA/OBNdba94wxQ4DP4ldWYjh5ZCE+j1G3pUhP1FAFT1wGL9wIw6bD11+F/pPcrkpEeqmODur/p7V2fPOzLa21a62158a3NPdlpfo5dkgeC5Ztw1p1W4r0GFs/grs/Dyueg5m3wEWPOY9AEhFxSUcH9RcbY54yxpRHlyeNMcXxLi4RzBpbyLodtawu3+12KSJyuKyFxffC32ZAqBEufx6O+5a6KEXEdR3tsrwPmAf0iy7PRI/1eDPGFAGwYPk2lysRkcPSuBv+fRU8+13n8UNffw0GHut2VSIiQMcDWR9r7X3W2lB0uR/oFU+yLcpKZsKAbI0jE+nOyj+Fv06DT56EaT+CS/4FafluVyUiEtPRQLbTGHOpMcYbXS4Fes2MqbPGFrK0tIotlfVulyIiB+vDR+GeaVBfCV+aC5+/ATx6+oaIJJaO/lb6Ks6UF2XAVuA84CtxqinhzBrrdFu+qG5Lke6jqQ6evgbmfgOKJ8PXX4chn3e7KhGRdnX0LssN1tozrbV9rLUF1tqzgB5/l2WzoX3SGdonTd2WIt3FjtXwt+nwwcNw4g1Oy1hGodtViYjs0+G021/faVV0A7PGFvHOugp21Ta5XYqI7M8n/4Z7Pg81W+GSJ+HkH4HX53ZVIiL7dTiBrFfdJz5zbBHhiOXlFeVulyIi7bEWXvsD/OtyKBjj3EU5vEc/4U1EepDDCWS9aqbU8f2zKMpMVrelSCKKROCFH8DCn8ER58FXnoOsXjFVooj0EPttxzfG1NB+8DJASlwqSlAej2Hm2EKeWLyJ+qYwKQGv2yWJCECoyRm4/8m/4Nhvwsxf6C5KEel29vtby1qbYa3NbGfJsNb2ukEZM8cU0RCM8Opn290uRUQAGmvg0QucMDb9pzDrlwpjItIt6TfXQZgyJJesFL+6LUUSwe7t8MAZsO5VmHMHnPBdPQJJRLqtXtfKdTj8Xg+njCpg4aflhMIRfF7lWRFX7FoPD50D1Vvgwkdh5Gy3KxIROSxKFAdp5thCquqDvLuuwu1SRHqnso/h7zOhbidc9rTCmIj0CApkB+nEEX1I8nnUbSnihvWvw32ngccHX50PA6e4XZGISKdQIDtIqQEfJ47ow4Ll27C2V838IeKu5fOcbsqMvnDFAigY5XZFIiKdRoHsEMwcU8jWqgY+3lzldikivcN7f4cnLoO+E+CrL2iOMRHpcRTIDsH00YV4DOq2FIk3a2HRr+G562H4TGfMWGqu21WJiHQ6BbJDkJMW4JjBuSxYts3tUkR6rkjYCWKLfgUTLoYLH4FAqttViYjEhQLZIZo1tojPynezdvtut0sR6XmCDfDPL8Pie+H46+Csv4DX73ZVIiJxo0B2iGaOLQJgwXK1kol0qoYqeOQ8+PQZmPUrmPEzTfgqIj2eAtkh6p+dwhH9MzWOTKQz1ZTBfV+AjW/BOX+Dz33T7YpERLqEAtlhmDWmiA82VrKtusHtUkS6v51rnAlfK9bCxU/A+PPdrkhEpMsokB2GWUc43ZYvqttS5PBsft8JY0274SvPwLBT3K5IRKRLxTWQGWNmG2NWGmNWG2Nu3M915xpjrDFmcjzr6WzDC9IpyUtVt6XI4VjzsvOQcH8qfHUB9J/kdkUiIl0uboHMGOMF7gBOBcYAFxljxrRzXQbwHeCdeNUSL8YYZo0t4q01O6mqD7pdjkj38/G/4JELIHuQM/t+/jC3KxIRcUU8W8iOAVZba9daa5uAfwBz2rnuZuA3QLcciDVzbCGhiGXRynK3SxHpXt6+C568AgYcA5c/D5l93a5IRMQ18Qxk/YFNrfZLo8dijDETgQHW2uf290bGmKuNMYuNMYu3b9/e+ZUehqMG5NAnI0ndliIdZS289DN44fsw6nS49N+Qku12VSIirnJtUL8xxgP8AfifA11rrb3HWjvZWju5T58+8S/uIHg8hhljClm0cjsNwbDb5YgktnAI5l0Lr/8BJn0FLngQ/MluVyUi4rp4BrLNwIBW+8XRY80ygCOARcaY9cCxwLzuNrAfnIeN1zWFeWP1DrdLEUlcTXXw+KXwwcPw+e/D6X8Cj9ftqkREEkI8A9l7wHBjzGBjTAC4EJjXfNJaW2WtzbfWllhrS4C3gTOttYvjWFNcHDc0n4wkn7otRfalrgIeOhtWvQCn3QrTfqjZ90VEWvHF642ttSFjzLXAfMAL3GutXWaM+Tmw2Fo7b//v0H0EfB6mjSrgpU/LCUcsXo/+ohGhoQpWLYAVz8BnL0EkCOffD2PPcrsyEZGEE7dABmCtfR54vs2xn+zj2pPiWUu8zRxbyLyPtrB4fQVThuS5XY6IO2rKYMVzsOJZWPeaE8LSCpxZ9yd+GfpPdLtCEZGEFNdA1pucNLKAgM/D/GXbFMikd9m5xnkQ+IpnofQ951jOYDj26zDqDCg+Gjx6KIiIyP4okHWS9CQfJwzLZ8HyMn58+miMxsdIT2UtbP0QPn3WaQ3b/qlzvO8EmPYjGPUFKBitMWIiIgdBgawTzRxTyMsrylm+tZqx/bLcLkek84RDsPHNlhBWXQrGA4OOh0m/dkJY9kC3qxQR6bYUyDrR9DGFmKc+ZsGybQpk0v0F653nTK54Dlb+B+orwJcMQ0+GaT+AEadCmrrnRUQ6gwJZJ8pPT2LyoBzmLyvjuzNGuF2OyMGr3wWr5jvjwVYvhGAdJGfBiNlOK9jQUyAp3e0qRUR6HAWy9lh7yONfZo0t4pbnPmXjzjoG5qV2cmEicVC9peXOyPWvQyQEGX3hyIudEFYyFbx+t6sUEenRFMjaKvvEebTLefdB7uCDfvnMMU4gW7C8jCunDolDgSKdIBKGj/8F794Nm5c4x/KGw3Hfcp4v2W+i7owUEelCCmRthRpg13q4dxZc+iQUjTuolw/MS2VUUQbzlymQSQKy1pmi4pVfwPYVUDAGTv4xjD4D+ox0uzoRkV5L/wRuq3gyXP4CeHxw32mw/o2DfotZY4tYvGEX22sa41CgyCGw1pkt/56T4IkvgY04s+Z//Q048f8pjImIuEyBrD0Fo+Cr8yGjyHn+3ornDurls8YWYS0s/HRbnAoUOQjr34D7ToVHznXulJzzF/jGWzD2bHVLiogkCP023pfsAU5LWdE4ePxSeP+hDr90dN8MinNS9LBxcdfmJc4/KO4/DSrWOQ/1vnYJHHUJeDVaQUQkkei38v6k5cFlT8MTlzkD/Wu3wwnfPeAdmMYYZo0t4qG3NrC7MUR6kn7M0oW2LXfGiK14FlJyYcbNcPSVENBdvyIiiUotZAeSlA4X/QOOOA8W/gzm3wSRyAFfNnNMIU3hCItWlndBkSI4z5R88iq48zhY+1846YfwnY/g+G8rjImIJDg13XSELwDn/BXS8uHtO6BuB8y5Y79zM00uySUvLcD8Zds4fXy/LixWep2qUvjvb+GDh8EbcALY8ddBaq7blYmISAcpkHWUxwOzfw1pfeDlm6GuAi54AAJp7V7u9Rimjy7kuY+30hgKk+TzdnHB0uPtLofX/gCL73Xumjz6Cpj6P87NKCIi0q2oy/JgGONMEXDGn2HNQnhwjhPM9mHm2EJ2N4Z4a83OLixSerz6XfDSz+DPE+Dde2D8+fDt9+G03ymMiYh0U2ohOxSTvgKpefCvK5zpBC79N2T13+uy44flkxbwMn/ZNk4aWdD1dUrP0rgb3rkT3rgdGqvgiHOdcWL5w9yuTEREDpNayA7V6DOcmfyrNsPfZ8L2VXtdkuz3ctLIAl5cvo1IxLpQpPQIwQZ46w6nRezlW6DkeGdC1/PuVRgTEekhFMgOx+CpcPlzEG5yHrVUumSvS2aOLWTH7kY+2LTLhQKlWwsHnfFhtx0F838IRUfAlQvhosecbRER6TEUyA5X3wlwxXxIzoQHzoDVC/c4PW1UAX6vYcEyzdovHRQJw0ePw/9Nhme/C1nF8OVnnDnxiie7XZ2IiMSBAllnyB3iPGopdwg8+kX4+F+xU5nJfj43NJ/5y8qwVt2W0o5ggzOr/nt/h3nfhv87Gp66GpIy4OIn4IoFMPhEt6sUEZE40qD+zpJR5HRfPnYRPHkl1O2EKV8DnElifzT3E1Zt283IogyXCxVXNdXBtk9g60ew5UNnvf1TiISc8yk50PdIOOXHMHqOnjUpItJLKJB1puQs547LJ6+A/3wPanfAtB/GAtmCZWUKZL1J424oW+qEruYAtmOlM2cYQGo+9DsSRsxyur77HQlZAw74aC4REel5FMg6mz8Zzn8Anr0OXv0t1JZT8IU/cNTAbOYvL+Nbpwx3u0KJh4Yq2Nocvj50wtfO1UC0mzq9yAldY8501n2PhMx+Cl8iIgIokMWH1wdn3g7pBfDa76FuJ6eN+iG/WLCO0l11FOfouYLdWl2F0/LV3OW49UOoWNtyPrO/E7jGne+0evWdoAlbRURkvxTI4sUYOOUnTrfU/B9waXUFf+ZyXly+jcuPH+x2dXIwmmrho3/A2kVOAKvc0HIue6ATuI68xAlhfSdAeh/XShURke5JgSzePvdNSMsnZe43eCp1K7cu/YUCWXexe7vzaKL3/uo8riinBPpPhMmXt4QvPcBbREQ6gQJZVxh/AaTkUvLYJfyg7DqqNg8jq/8It6uSfdm5Bt76P/jwUQg1wsjT4Phvw8Bj3a5MRER6KAWyrjJ8OhtO/wd5T19K4MFT4fKnNdt6otn0Hrz5Z/j0WfAGYMKFcNy3IF83YoiISHxpkqMuNPSok/hm4JfUhwzcdxqsf8PtkiQSgZX/gXtnw9+nw7pXYer1cN3HcOZtCmMiItIlFMi6kDGGkeMmc07j/xJJL4CHzoalT0A45HZpvU+oEd5/EP4yBR67EKpKYfav4bvLnZsxMgrdrlBERHoRBbIuNnNMIetDubxy3INQNA7+fRX8cQzMv8m5g0+PV4qv+l3OVCR/GgfzvgW+JDjnb/DtD+DYb0BSutsViohIL6QxZF3smMG5ZKf6eXZ1E6dc/h9Y9QIsfRzeudsZSN5ntHMTwPgLnIdKS+eo3ARv3wnvPwBNu2HoyXD23TDkJE3OKiIirlMg62I+r4dTRhXy4vIygmY8/jFnOrO311XAsqecLsyFP4OFP4eSE2D8F53zyVlul949lX0Mb9wGnzzp7I87zxmoXzTO3bpERERaMbabdZFNnjzZLl682O0yDsv8ZWV87aElPHzFFE4Ynr/3BRXrnGC29HGoWAO+ZBh5qhPOhk0Hr7/ri+5OrIW1rzhBbO0rEEiHiV92uiSzB7hdnYiI9CLGmCXW2skHuk4tZC44cXgfkv0eFiwvaz+Q5Q6Gk74Pn/8ebF7iBLNPnnRa0FLzYOw5zpQM/Sepu621cBCWzXWmrij7GNIL4ZT/dSZyTclxuzoREZF9UguZS7720GI+2lTFmzeejMfTgVAVDsLqhbD0H7DieQg3Qu5Qp9Vs/AVOiOutGnc7d0y+/Reo2gT5I5xuyfFfdAbti4iIuEQtZAlu5pgi5i/bxtLNVRw5IPvAL/D6YeRsZ2moguXznJazRb90lgFTnAAy9uyufZyPtdBYAzVbo0sZ7C4HG4m23hkwHmfbeKL7zdu0Oe7p4GtaXVe2FN77OzRUwsDPwWm/g+GzwKMbiEVEpPtQIHPJKaML8HoMC5aVdSyQtZacBRO/5CyVm+Djfzrh7Lnr4T/fhxGznHA2YtbhtRCFGp2AVVMGNVucdXV03RzAqrdCsPbQP+OwGRh9Ohz3HRhwtIt1iIiIHDp1Wbro4r++zfodtTz/nalkpwYO782sdVqLPnrcCWi15U5wG3u2E84GHNvSahSJQN2OVuGqbdiKHqvbuffneAOQ0ddZMqPrjCLI6OesM/tBWh/w+ADrtJTZ6Bob3bZ7ntvndR14TUquU4eIiEgC6miXpQKZi95bX8Elf32H8cVZPHzlFJL93s5543AI1i1ywtmKZyFYB9kDIa0g2qVYBpG2TwcwkF6wd7hqu5+SoxsJREREOkiBrJt4bulWrn3sfWaOKeQvl0zC25EB/gejcbcTyj75t3MjQHPr1h4tXH2dMKbpNERERDqVBvV3E18Y35fymjH87Jnl/HTeMn4+ZyymM1ugktKdKTImXNh57ykiIiKdSoEsAVx+/GDKqhq4+9W1FGUlc820YW6XJCIiIl1IgSxBfH/2KLZVN/C7+SspzEzmvEl6jqWIiEhvoUCWIDwew2/Pm8D23Y3c+ORS8tMDnDSywO2yREREpAto9swEEvB5uOvSSYwozOCbj7zPx6VVbpckIiIiXUCBLMFkJPu5//KjyUkNcPn977Jhp5uTroqIiEhXUCBLQAWZyTzw1f/f3p3HR1Xfff9/fTMz2fcECNkAhRKWCMgiiGAFq4hUEYWIuPfW36XWuly/Vlwul6u2tbV379brdqvKVfFSFLeql4oCopQLBAJEWUzZzcKWjZAA2Wa+9x8zZIEEAiaZZPJ+Ph7zmDPnnDnzmcNk5s35fs/3jKHOY7lp3hpKKqv9XZKIiIi0IwWyTqp/z0heuWkUe8uruPXVbI7UHD+Qq4iIiAQKBbJObGSfeP5j9gg2Fhzk529soM7t8XdJIiIi0g4UyDq5S4Yk8evpQ/ki9wAPv7+JrnZlBRERETk1DXvRBcw5rw/7y6t45ovt9IoJ5f6f/MjfJYmIiEgbUiDrIu77yY/YW17FM0u3kRQdynXnpfu7JBEREWkjCmRdhDGG387IpKiymkf+vuliGbUAACAASURBVJGeUSFcPLiXv8sSERGRNqA+ZF2IyxHEc3POJTMlhp8vWM/6vDJ/lyQiIiJtQIGsiwkPdvLKzaNJig7lZ39by46iSn+XJCIiIj+QAlkXlBgZwqu3jiHIGG6at4YDFVX+LklERER+AAWyLqpPQgT/ectoSg/XcMt/rqWiqtbfJYmIiMgZUiDrws5JjeXZOeeSu6+CO/5rPTV1GjhWRESkK1Ig6+IuGtiTp2ZksmJ7MQ+8+y0ejwaOFRER6Wo07EUAmDkqjf2Hqvjj51vpFR3K3Msy/F2SiIiInAYFsgBx10X92Xeoihe+2kFSdAg3j+/n75JERESklRTIAoQxhieuGMqBQ9U88d9b6BkdytTM3v4uS0RERFpBfcgCiCPI8MzsEZybHse9b+WwemeJv0sSERGRVlAgCzChLgcv3ziKtLgwbpufzdb9Ff4uSURERE5BgSwAxUUE8+qtYwh1Obhp3hr2lh/1d0kiIiJyEgpkASo1Lpy/3TKGiqo6bp63lvKjGjhWRESks1IgC2CDk6N58YaR7Cyu5Pb52VTVuv1dkoiIiDRDgSzAje+fyB9nDmP1rlLmvLxaFyMXERHphBTIuoErh6fwl2uHs/1AJZf95R889+V26ty6zJKIiEhn0a6BzBgzxRjzT2PMdmPM3GaW32+M2WKM+dYYs9QY06c96+nOrhyewuL7JzJpYE/+sOifTH/uf9iy55C/yxIRERHaMZAZYxzAs8BlwGBgtjFm8HGrbQBGWWvPAd4B/tBe9Qj0jArlhRtG8tycc9lXXsUV/3cF//vzf1Jdp75lIiIi/tSeR8jGANuttTuttTXAm8CVjVew1i6z1h7xPfwaSG3HesRnamZvFt93IVcMS+Y/vtjOtGdWsD6vzN9liYiIdFvtGchSgPxGjwt881ryM+DT5hYYY243xmQbY7KLiorasMTuKy4imD9lDec/bx5NZXUdVz+/kl//9xaO1uhomYiISEfrFJ36jTHXA6OAp5tbbq39q7V2lLV2VI8ePTq2uAB3UUZPPr9vIteNSeeVFbu49M/LWbmj2N9liYiIdCvtGcgKgbRGj1N985owxlwMPAxcYa2tbsd6pAVRoS5+c1UmC24bizFw3UurefC9jRyq0mCyIiIiHaE9A9laYIAxpp8xJhi4Fviw8QrGmBHAi3jD2IF2rEVaYdzZCSy6ZyK3TejHW2vzuORPy/kid7+/yxIREQl47RbIrLV1wM+Bz4DvgIXW2s3GmH83xlzhW+1pIBJ42xiTY4z5sIXNSQcJC3bw8OWDee/O8USHObn1b9nc++YGSg/X+Ls0ERGRgGWstf6u4bSMGjXKZmdn+7uMbqG6zs2zy3bw3LLtxIS5eOLKIVye2RtjjL9LExER6RKMMeustaNOtV6n6NQvnVOI08H9P/kRH919AcmxYfz8jQ38f6+t48ChKn+XJiIiElAUyOSUBvWO5v07z2fuZRl8tbWIi//0FQuz8+lqR1dFREQ6KwUyaRWnI4h/ufBsPr1nAgOTovjVO99y47w1FJQdOfWTRURE5KQUyOS0nNUjkrduH8e/XzmE9d+Xccn/Wc78VbvxeHS0TERE5EwpkMlpCwoy3DiuL5/dN5GRfeJ49IPNZP11FTuLKv1dmoiISJekQCZnLDUunPm3juHpa87hn/sqmPKXf/DCVzuoc3v8XZqIiEiXokAmP4gxhpmj0lhy/4VcNLAHT32ay1XPrWTRpn3UKpiJiIi0isYhkzZjreWTjft48uMt7C2vokdUCLNGpXLt6HTS4sP9XZ6IiEiHa+04ZApk0ubq3B6+2lrEgjV5fJF7AI+FCQMSmT0mnYsH9SLYqQOzIiLSPSiQSaewt/woC9cWsDA7n8KDR0mMDOaakWlcOzqNvokR/i5PRESkXSmQSafi9liWbytiweo8luYewO2xnH92ArPHpHPJkF6EOB3+LlFERKTNKZBJp7X/UBVvZ+fz5tp8CsqOEh8RzDUjU7l2dBpn9Yj0d3kiIiJtRoFMOj2Px7JiezEL1uSxeMt+6jyW8/rFc9156Vw6JIlQl46aiYhI16ZAJl3KgYoq3llXwJtr8skrPUJsuIsZI1KZPSaNAb2i/F2eiIjIGVEgky7J47Gs2lnCG2vy+HzzPmrdltF945g9Jp2pmb111ExERLoUBTLp8oorq3l3XQEL1uSxu+QI0aFOZpybyuwx6QxM0lEzERHp/BTIJGBYa/l6ZykL1uSxaNM+atwezk2PZfaYdKadk0xYsI6aiYhI56RAJgGp9HAN7633HjXbUXSYqBAn00ekMHtMOoOTo/1dnoiISBMKZBLQrLWs2VXKm2vz+XjjXmrqPAxLjWH2mHR+OiyZiBCnv0sUERHpXoGstraWgoICqqqq/FSVAISGhpKamorL5erQ1z14pIb3NxSyYE0eW/dXEhHs4MoRKcwenU5makyH1iIiItJYtwpku3btIioqioSEBIwxfqqse7PWUlJSQkVFBf369fNbDevzyliwJp///nYPVbUehqZEM3tMOlcMSyYqtGODooiISLcKZN999x0ZGRkKY35mrSU3N5dBgwb5uxTKj9byQU4hb6zOI3dfBeHBDn56TjKzz0tnWGqMPisiItIhWhvIAqajjX5g/a8z/RvEhLm4cVxfbhjbh28KylmwOo8Pv9nDW9n5ZCRFcd156Vw5PIWYMB01ExER/wuYQCbSHGMMw9NiGZ4WyyPTBvHhN3t4Y3Uej36wmd9+8h3Tzklm9pg0zk2P61SBUkREuhcFMuk2okJdzDmvD3PO68PGgnLeWJPHhzmFvLOugB/1imT2mHSuGpFCbHiwv0sVEZFuJsjfBQSCgwcP8txzz53286ZOncrBgwdPus6jjz7KkiVLzrS0ZkVGRrbp9rqizNQYfjcjkzUPX8xTMzIJczl44qMtjPntUu57K4c1u0rpav0rRUSk6wqYTv3+7Ei+e/dupk2bxqZNm5rMr6urw+nsfAchIyMjqaysbJdt+/vf4ofYvKecN9fk8/cNhVRU13F2jwhmj0lnxrmpxEfoqJmIiJy+btep/5gnPtrMlj2H2nSbg5OjeeynQ1pcPnfuXHbs2MHw4cNxuVyEhoYSFxdHbm4uW7duZfr06eTn51NVVcU999zD7bffDkDfvn3Jzs6msrKSyy67jAsuuICVK1eSkpLCBx98QFhYGDfffDPTpk3jmmuuoW/fvtx000189NFH1NbW8vbbb5ORkUFRURHXXXcde/bsYdy4cSxevJh169aRmJh40vdlreVXv/oVn376KcYYHnnkEbKysti7dy9ZWVkcOnSIuro6nn/+ec4//3x+9rOfkZ2djTGGW2+9lfvuu69N97O/DUmO4dfTY3hwagYff7uXBWvyePLj7/jDon9y6dAkpg9P5ryzEojUoLMiItLG9MvSBp566ik2bdpETk4OX375JZdffjmbNm2qH49r3rx5xMfHc/ToUUaPHs3VV19NQkJCk21s27aNBQsW8NJLLzFr1izeffddrr/++hNeKzExkfXr1/Pcc8/xxz/+kZdffpknnniCSZMm8eCDD7Jo0SJeeeWVVtX93nvvkZOTwzfffENxcTGjR49m4sSJvPHGG1x66aU8/PDDuN1ujhw5Qk5ODoWFhfVHAU/V1NqVhQc7mTkqjZmj0vjnvgoWrMnjvfUFfPTNHpxBhhHpsVzQvwcXDEhkWGoMToda/kVE5IcJuEB2siNZHWXMmDFNBkd95plneP/99wHIz89n27ZtJwSyfv36MXz4cABGjhzJ7t27m932jBkz6td57733AFixYkX99qdMmUJcXFyr6lyxYgWzZ8/G4XDQq1cvLrzwQtauXcvo0aO59dZbqa2tZfr06QwfPpyzzjqLnTt3cvfdd3P55ZdzySWXtH6HdGEDk6J4/IohPDg1g3W7y/jH9mJWbCvmz0u38n+WbCUqxMnYsxOYMCCRC/on0i8xQmdriojIaQu4QNYZRERE1E9/+eWXLFmyhFWrVhEeHs6Pf/zjZi/xFBISUj/tcDg4evRos9s+tp7D4aCurq6NK/eaOHEiy5cv5+OPP+bmm2/m/vvv58Ybb+Sbb77hs88+44UXXmDhwoXMmzevXV6/MwpxOji/fyLn90/kgSlQdriGlTtKWLG9iH9sK2bxlv0AJMeEcsGARC4Y0IPxZyeQEBlyii2LiIgokLWJqKgoKioqml1WXl5OXFwc4eHh5Obm8vXXX7f5648fP56FCxfywAMP8Pnnn1NWVtaq502YMIEXX3yRm266idLSUpYvX87TTz/N999/T2pqKrfddhvV1dWsX7+eqVOnEhwczNVXX83AgQObbU7tTuIigrn8nN5cfk5vrLXklR7hH9u8R88WbdrHwuwCAAb3jmbCgETG909kTL94Ql0OP1cuIiKdkQJZG0hISGD8+PEMHTqUsLAwevXqVb9sypQpvPDCCwwaNIiBAwcyduzYNn/9xx57jNmzZ/Paa68xbtw4kpKSiIqKOuXzrrrqKlatWsWwYcMwxvCHP/yBpKQkXn31VZ5++mlcLheRkZHMnz+fwsJCbrnlFjweDwC/+93v2vx9dFXGGPokRNAnIYLrx/bB7bFsLCxnxbYiVmwvZt7/7OLF5TsJdgYxum8c4/snMqF/D4YkRxMUpOZNERHRsBcBobq6GofDgdPpZNWqVdxxxx3k5OT4pZbu/m/RnCM1dazeVcqKbcX8z/Zicvd5j6bGhbs4/+xEbxNn/0TS4sP9XKmIiLS1bjvsRXeUl5fHrFmz8Hg8BAcH89JLL/m7JGkkPNjJRQN7ctHAngAcqKjif7YXs2Kbtw/axxv3AtAnIZwL+icyYUAiw9JiSYoO1QkCIiLdhAJZABgwYAAbNmxoMq+kpITJkyefsO7SpUtPOMNTOlbPqFCuGpHKVSNSsday/UAlK3xnb/59QyGvr84DvEfQBidHM7h3tO8+hrN7RGiYDRGRAKRAFqASEhL81mwprWeMYUCvKAb0iuKW8f2odXv4tuAgm/ccYsueQ2zZe4hXV31PTZ23716wM4iBvaIaQlpyNBlJUUSFuvz8TkRE5IdQIBPpRFyOIEb2iWdkn/j6eXVuDzuLD9cHtC17DvH5ln28lZ1fv06fhHBvSGsU1NTkKSLSdSiQiXRyTkcQP+oVxY96RTF9RArgvezV/kPVbNlb3iSofbppX/3z4iOCGzV3eu/PSlSTp4hIZ6RAJtIFGWNIigklKSaUSRkNw6xUVteRu7choG3Ze4i/rdzdpMkzI6mhyXNIcjRDkmM0PpqIiJ8pkIkEkMgQJ6P6xjOq78mbPD/bvI8313qbPJ1BhozeUYxIi2N4Wiwj0mN1CSgRkQ6mQOYHkZGRVFZWNrts9+7dTJs2rf4i3iI/1MmaPDcWlpOTX0ZO/kHe31DIa19/D0BMmIvhabHeW3osI9JiiQ0P9ufbEBEJaApkIt1Q4ybPnwz2Nnm6PZYdRZVsyPMGtA15B/mPL7bh8Y0d3S8xghH1AS2OjN5RuNQfTUSkTQReIPt0Luzb2LbbTMqEy55qcfHcuXNJS0vjrrvuAuDxxx/H6XSybNkyysrKqK2t5cknn+TKK688rZetqqrijjvuIDs7G6fTyZ/+9CcuuugiNm/ezC233EJNTQ0ej4d3332X5ORkZs2aRUFBAW63m3/7t38jKyvrB71t6V4cQab+SFrW6HTA2yft24KD5OQfJCfvIP/YXsx7GwoBCHEGMTQlpiGkpceRHKMzO0VEzkTgBTI/yMrK4t57760PZAsXLuSzzz7jF7/4BdHR0RQXFzN27FiuuOKK0/qxevbZZzHGsHHjRnJzc7nkkkvYunUrL7zwAvfccw9z5syhpqYGt9vNJ598QnJyMh9//DHgvai5yA8VGeLk/LMTOf/sRMDb1LmnvMp7FC3vIBvyD/La19/z8opdAPSICqkPaMPTYjknNZbIEH3NiIicSuB9U57kSFZ7GTFiBAcOHGDPnj0UFRURFxdHUlIS9913H8uXLycoKIjCwkL2799PUlJSq7e7YsUK7r77bgAyMjLo06cPW7duZdy4cfzmN7+hoKCAGTNmMGDAADIzM/nXf/1XHnjgAaZNm8aECRPa6+1KN2aMISU2jJTYMKadkwxATZ2H3H2H6o+ibcg/yOdb9gMQZOBHvaIYke4NZ5kpMfyoVxTBTjV1iog0FniBzE9mzpzJO++8w759+8jKyuL111+nqKiIdevW4XK56Nu3L1VVVW3yWtdddx3nnXceH3/8MVOnTuXFF19k0qRJrF+/nk8++YRHHnmEyZMn8+ijj7bJ64mcTLAziHNSvYHrxnHeeWWHa8gpaAhon2zcx4I13rM6XQ7DwKQoMlNiGJoSw9DkGAYmRWnoDRHp1hTI2khWVha33XYbxcXFfPXVVyxcuJCePXvicrlYtmwZ33///Wlvc8KECbz++utMmjSJrVu3kpeXx8CBA9m5cydnnXUWv/jFL8jLy+Pbb78lIyOD+Ph4rr/+emJjY3n55Zfb4V2KtE5cRHCTC6p7PJb8siNsLCxnY2E5mwsPNQlpziDvJaQyU6K9IS0lhsG9oxXSRKTbUCBrI0OGDKGiooKUlBR69+7NnDlz+OlPf0pmZiajRo0iIyPjtLd55513cscdd5CZmYnT6eRvf/sbISEhLFy4kNdeew2Xy0VSUhIPPfQQa9eu5Ze//CVBQUG4XC6ef/75dniXImcmKMjQJyGCPgkR9U2d1loKyo6yyRfSNu05xJLvDrAwuwDwnmTQv0ekL6BFk5kSw+DkaMKD9bUlIoHHWGv9XcNpGTVqlM3Ozm4y77vvvmPQoEF+qkga07+F/BDWWvaWV/mOopX7jqgdoriyGgBj4OwekWSmxDAk2RvShqTE6MQBEem0jDHrrLWjTrWevsVEpNMwxpAcG0ZybBiXDvGeAGOt5UBFNRsLytm0p5xNheWs3FHM+77hN4yBfgkR9UfShqbEcFZiJD2jQggK0hAcItI1KJD5ycaNG7nhhhuazAsJCWH16tV+qkikczLG0Cs6lF6DQ7l4cMN1Ow9UVLG58FB9k+e678v48Js99cuDHUEkx4aSGhdOalwYqXFhpMSF1T/uGRWKQ4FNRDoJBTI/yczMJCcnx99liHRZPaNC6ZkRykUZPevnlVRWs2XvIb4vOUJB2VEKyrz3S747UN/seYzL4T0alxoXRmpsuC+sNQS2XtEKbCLScRTIRCRgJESGMGFADyYMOHFZVa2bwoNHmwS1Y9Nf/PMARRVNA5szyBvYUmKbBrXUuDBS48PpFRWCU5eOEpE2okAmIt1CqMvB2T0iObtHZLPLq2rd7KkPbI1D2xGWbyti/6Gmgc0RZEiKDqVndAg9o0LoERVCj8hQ732jW2JkMCFODd8hIienQCYigjewndUjkrNOEtj2llc1CWqFZUcprqxhV/Fh1uwqpexIbbPPjQ130SOyUVCLbBraekZ5g1xsmEsnIoh0UwpkIiKtEOpy0C8xgn6JES2uU1PnoeRwNUUVTW8Hjk1XVrMh7yAHKqqoqvWc8HxnkCGxheDWKzqU1Lgw0uLDiQlztedbFRE/UCBrAwcPHuSNN97gzjvvPK3nTZ06lTfeeIPY2NgW13n00UeZOHEiF1988Q8tU0TaWbAziN4xYfSOCTvpetZaDte4GwW2qqYhrrKa/Yeq2FRYTnFlNZ7jhouMCnWS5uvTlhbvu48LJzXe29dN47KJdD0BNzDs79f8ntzS3DZ9zYz4DB4Y80CLy3fv3s20adPYtGlTk/l1dXU4nd3ri1EDw4q0LbfHUnq4hn2+5tL8Rick5Jd6p4/Wups8Jy7cRWpcOGm+gJbmOykhLT6MlNhwwoLVp02ko2hg2A40d+5cduzYwfDhw3G5XISGhhIXF0dubi5bt25l+vTp5OfnU1VVxT333MPtt98OQN++fcnOzqayspLLLruMCy64gJUrV5KSksIHH3xAWFgYN998M9OmTeOaa66hb9++3HTTTXz00UfU1tby9ttvk5GRQVFREddddx179uxh3LhxLF68mHXr1pGYmNhsvS3Vs2jRIh566CHcbjeJiYksXbqUyspK7r77brKzszHG8Nhjj3H11Vd32L4V6e4cQaa+2TIzNeaE5dZaSg7X1Pdryy/13ZcdJXdfBUu+O0BNXdPm0cTIkPozRpscYfON1aaTEEQ6XsAFspMdyWovTz31FJs2bSInJ4cvv/ySyy+/nE2bNtGvXz8A5s2bR3x8PEePHmX06NFcffXVJCQkNNnGtm3bWLBgAS+99BKzZs3i3Xff5frrrz/htRITE1m/fj3PPfccf/zjH3n55Zd54oknmDRpEg8++CCLFi3ilVdeOWm9zdXj8Xi47bbbWL58Of369aO0tBSAX//618TExLBx40YAysrK2mKXiUgbMcbb7ywxMoThaSd2f/B4LMWV1eQ3OnP02JG1TYXlfLZ5H7Vu22h70CsqlPT4cNLivUfV0uPD6289okIwRiceiLS1gAtkncGYMWPqwxjAM888w/vvvw9Afn4+27ZtOyGQ9evXj+HDhwMwcuRIdu/e3ey2Z8yYUb/Oe++9B8CKFSvqtz9lyhTi4uJOWl9z9RQVFTFx4sT6uuPj4wFYsmQJb775Zv1zT7VtEelcgoIMPaND6Rkdysg+J/79uj2WAxVVDUfWSo+SX3aEvNIjrNpRzHsbqmjcsyXEGURao4CWGucLbAnhpMWFE6H+ayJnRH857SAiouEsrC+//JIlS5awatUqwsPD+fGPf0xVVdUJzwkJCamfdjgcHD16tNltH1vP4XBQV1d32rW1th4R6R4cQab+RIQx/eJPWF5d56aw7Ch5pUfIL/U2heaVeAPb2l2lVFQ3/R5KiAhuEtjS470nG6THh9M7JkxXPxBpgQJZG4iKiqKioqLZZeXl5cTFxREeHk5ubi5ff/11m7/++PHjWbhwIQ888ACff/75SZsVW6pn7Nix3Hnnnezatau+yTI+Pp6f/OQnPPvss/z5z38GvE2WOkom0n2EOFsen81aS/nRWvJKj9Tf8ku9R9ly8g/y8ca9uBudIupyGFJiw3xNoeGkxIYRHuzA6Qgi2GFwOYJ8t8bTxz/2TTt900FNpzWOm3RVCmRtICEhgfHjxzN06FDCwsLo1avhAshTpkzhhRdeYNCgQQwcOJCxY8e2+es/9thjzJ49m9dee41x48aRlJREVFRUs+u2VE+PHj3461//yowZM/B4PPTs2ZPFixfzyCOPcNdddzF06FAcDgePPfZYfbOpiHRvxhhiw4OJDQ/mnNQT+6/VuT3sLa+qD2qNQ9unG/e2OJDuD+EIMvWhLdgRhNM3HR7sICEihETf1RMSI73jvCVGBdf3wUvQVRXEjwJu2IvuqLq6GofDgdPpZNWqVdxxxx1+u3B5d/+3EJHWO1rjprrOTY3bQ63bUuf2UOv2UFNnqXV7qPO0PF1bZ33P81Dnbjpd6/ZQc9z0kWo3xZXVvlsNldXNd/mIDnX6QpsvsPnC27F59WEuKoRQl8KbnJqGvehG8vLymDVrFh6Ph+DgYF566SV/lyQickphwQ6/jYlWVesdmPdYQCuurKa40eOiymq+23eIoopqKqqaD2+RIc6GwNboaFtUqIsQZxAhziCCnUGEOB31j0NcQQQ7HIS4mi4P9i13BhmdxdpNKZAFgAEDBrBhw4Ym80pKSpg8efIJ6y5duvSEMzxFRLqbUJejvi/bqVTVuik5XNMosPlCW6PH24sq+XpXNQd/YDNskOGEkOYNbo5GAa4h5DXpUxfU/LQzyBDsbL5PntNhCG7FdKjLQUSwQ2GxHSmQBaiEhAS/NVuKiASSUJeDlNgwUmJPfkks8F7P9FhTbHWdx3dzU+ObbnrvbjJ9/DqNt9GwzE1ldR2lh2uorjuxyba2zkOtx54wGHBbCHEG1TfbJjS5b2jKTfDdx4UH64za06RAJiIi0kaCfUexwL8XgLfW4vZY6jy+sFbn8U77Qlytr39dbaP+ezXNTHsDn4cjNW5KD3ubcksqa9h/qIrNe8opqayh7viLreI90hcfEew7kcJ37wtsPRoFt2P3re2PZ633/VTVeqiudVNV66Gqzk2Vb/po7bFpN9WNlh2tabpeda2bqjo3Vw5P4dIhSW29+89IuwYyY8wU4C+AA3jZWvvUcctDgPnASKAEyLLW7m7PmkRERAKdMQanw+B00K4nHxwb+uRYP7yS+vtqiiprKPE16X5TdpDiimoO17ib3c6x/ngJkSGEOIO8Iao+dLmpqvPUB61m8l+rOIMMYS4HIS4Hoa4gQl0OytvhTN8z1W6BzBjjAJ4FfgIUAGuNMR9aa7c0Wu1nQJm1tr8x5lrg90BWe9UkIiIibafx0Cf9e544Vt3xjtY0nO1aH94Oe/vjHeunV1PnITzYSXxEQ3AKdQUR6nQ0TLscjW4tLWs07QzC6QjqgD1y5trzCNkYYLu1dieAMeZN4EqgcSC7EnjcN/0O8H+NMcZ2tbE4RERE5JTCglt/MkV3055xMQXIb/S4wDev2XWstXVAORDwpwBGRp76fxEiIiLSfXSJTv3GmNuB2wHS09NPuu6+3/6W6u9y2/T1QwZlkPTQQ226TREREZFj2vMIWSGQ1uhxqm9es+sYY5xADN7O/U1Ya/9qrR1lrR3Vo0ePdir3zM2dO5dnn322/vHjjz/Ok08+yeTJkzn33HPJzMzkgw8+aNW2KisrW3ze/PnzOeeccxg2bBg33HADAPv37+eqq65i2LBhDBs2jJUrV7btmxMREZF2126XTvIFrK3AZLzBay1wnbV2c6N17gIyrbX/4uvUP8NaO+tk2+2Ml07asGED9957L1999RUAgwcP5rPPPiMmJobo6GiKi4sZO3Ys27ZtwxhDZGQklZWVzW6rrq6OI0eOnPC8LVu2cNVVV7Fy5UoSExPrL/6dlZXFuHHjuPfe9uV3jgAACj5JREFUe3G73VRWVhITE9ORb78Jf/9biIiIdCZ+v3SStbbOGPNz4DO8w17Ms9ZuNsb8O5Btrf0QeAV4zRizHSgFrm2vetrTiBEjOHDgAHv27KGoqIi4uDiSkpK47777WL58OUFBQRQWFrJ//36Skk4+3om1loceeuiE533xxRfMnDmTxMREAOLj4wH44osvmD9/PgAOh8OvYUxERETOTLv2IbPWfgJ8cty8RxtNVwEz27OGjjJz5kzeeecd9u3bR1ZWFq+//jpFRUWsW7cOl8tF3759qaqqOuV2zvR5IiIi0nV17kE5upCsrCzefPNN3nnnHWbOnEl5eTk9e/bE5XKxbNkyvv/++1Ztp6XnTZo0ibfffpuSEm8Xu9LSUgAmT57M888/D4Db7aa8vLwd3p2IiIi0JwWyNjJkyBAqKipISUmhd+/ezJkzh+zsbDIzM5k/fz4ZGRmt2k5LzxsyZAgPP/wwF154IcOGDeP+++8H4C9/+QvLli0jMzOTkSNHsmXLlpNtXkRERDqhduvU3146Y6d+aaB/CxERkQat7dSvI2QiIiIiftYlBoYNRBs3bqwfS+yYkJAQVq9e7aeKRERExF8CJpBZazHG+LuMVsvMzCQnJ8ffZbSprtb8LSIi0lkERJNlaGgoJSUlCgR+ZK2lpKSE0NBQf5ciIiLS5QTEEbLU1FQKCgooKirydyndWmhoKKmpqf4uQ0REpMsJiEDmcrno16+fv8sQEREROSMB0WQpIiIi0pUpkImIiIj4mQKZiIiIiJ91uZH6jTFFQOsuDBm4EoFifxfRSWhfNNC+aKB94aX90ED7ooH2RYOO2Bd9rLU9TrVSlwtkAsaY7NZchqE70L5ooH3RQPvCS/uhgfZFA+2LBp1pX6jJUkRERMTPFMhERERE/EyBrGv6q78L6ES0LxpoXzTQvvDSfmigfdFA+6JBp9kX6kMmIiIi4mc6QiYiIiLiZwpkIiIiIn6mQNZJGWPSjDHLjDFbjDGbjTH3NLPOj40x5caYHN/tUX/U2hGMMbuNMRt97zO7meXGGPOMMWa7MeZbY8y5/qizvRljBjb6984xxhwyxtx73DoB+7kwxswzxhwwxmxqNC/eGLPYGLPNdx/XwnNv8q2zzRhzU8dV3fZa2A9PG2NyfZ//940xsS0896R/S11NC/vicWNMYaO/gaktPHeKMeafvu+NuR1XdftoYV+81Wg/7DbG5LTw3ID5XLT0+9npvyustbp1whvQGzjXNx0FbAUGH7fOj4H/9netHbQ/dgOJJ1k+FfgUMMBYYLW/a+6AfeIA9uEddLBbfC6AicC5wKZG8/4AzPVNzwV+38zz4oGdvvs433Scv99PG++HSwCnb/r3ze0H37KT/i11tVsL++Jx4P8/xfMcwA7gLCAY+Ob479iudmtuXxy3/H8Djwb656Kl38/O/l2hI2SdlLV2r7V2vW+6AvgOSPFvVZ3alcB86/U1EGuM6e3votrZZGCHtbbbXLnCWrscKD1u9pXAq77pV4HpzTz1UmCxtbbUWlsGLAamtFuh7ay5/WCt/dxaW+d7+DWQ2uGF+UELn4nWGANst9butNbWAG/i/Sx1WSfbF8YYA8wCFnRoUX5wkt/PTv1doUDWBRhj+gIjgNXNLB5njPnGGPOpMWZIhxbWsSzwuTFmnTHm9maWpwD5jR4XEPgB9lpa/nLtLp8LgF7W2r2+6X1Ar2bW6W6fj1vxHjFuzqn+lgLFz33Nt/NaaJrqbp+JCcB+a+22FpYH5OfiuN/PTv1doUDWyRljIoF3gXuttYeOW7web3PVMOA/gL93dH0d6AJr7bnAZcBdxpiJ/i7In4wxwcAVwNvNLO5On4smrLfNoVuP5WOMeRioA15vYZXu8Lf0PHA2MBzYi7eprrubzcmPjgXc5+Jkv5+d8btCgawTM8a48H6YXrfWvnf8cmvtIWttpW/6E8BljEns4DI7hLW20Hd/AHgfb3NDY4VAWqPHqb55geoyYL21dv/xC7rT58Jn/7Hmad/9gWbW6RafD2PMzcA0YI7vB+cErfhb6vKstfuttW5rrQd4iebfY7f4TAAYY5zADOCtltYJtM9FC7+fnfq7QoGsk/K1978CfGet/VML6yT51sMYMwbvv2dJx1XZMYwxEcaYqGPTeDsvbzputQ+BG31nW44Fyhsdmg5ELf5vt7t8Lhr5EDh2JtRNwAfNrPMZcIkxJs7XfHWJb17AMMZMAX4FXGGtPdLCOq35W+ryjus/ehXNv8e1wABjTD/fEedr8X6WAtHFQK61tqC5hYH2uTjJ72fn/q7w99kQurV4lsgFeA+nfgvk+G5TgX8B/sW3zs+BzXjPDvoaON/fdbfTvjjL9x6/8b3fh33zG+8LAzyL96ypjcAof9fdjvsjAm/Aimk0r1t8LvCG0L1ALd6+HT8DEoClwDZgCRDvW3cU8HKj594KbPfdbvH3e2mH/bAdb9+XY98XL/jWTQY+8U03+7fUlW8t7IvXfN8D3+L9Ee59/L7wPZ6K9wy8HYG6L3zz/3bs+6HRugH7uTjJ72en/q7QpZNERERE/ExNliIiIiJ+pkAmIiIi4mcKZCIiIiJ+pkAmIiIi4mcKZCIiIiJ+pkAmIl2KMcZtjMlpdJvbhtvua4z5weMvGWOWGWNCjTF/NsaMa4vaRCSwOf1dgIjIaTpqrR3u7yJaYowJAzzW2ipjzGjgl/6uSUQ6Px0hE5GAYIzZbYz5gzFmozFmjTGmv29+X2PMF74LTS81xqT75vcyxrzvuwj7N8aY832bchhjXjLGbDbGfO4LWBhjfmGM2eLbzpst1LAM74CkQ40xG4FMYK0xZmq77wAR6dI0MKyIdCnGGDfe0HPM76y1bxljdgMvWWt/Y4y5EZhlrZ1mjPkIeMda+6ox5la8lxaabox5C1hlrf2zMcYBRAJxeEfnHmWtzTHGLAQ+tNb+lzFmD9DPWlttjIm11h5sob5fAjuBYmCatVZHyETklBTIRKRLMcZUWmsjm5m/G5hkrd3pu7DwPmttgjGmGO+lc2p98/daaxONMUVAqrW2utE2+gKLrbUDfI8fAFzW2ieNMYuASuDvwN+t7wLuzdSxAPg3vBeAL7PW/lfbvXsRCVTqQyYigcS2MH06qhtNu4Ew3/TlwETgp8DDxphMa23dsRWNMf8L73VE+wODgHRgvzHmMmvtnDOsRUS6CfUhE5FAktXofpVveiVwrW96DvAP3/RS4A4AY4zDGBPT0kaNMUFAmrV2GfAAEIO3ibOetfZl4BLgC99JB9uttYMUxkSkNXSETES6mjBjTE6jx4ustceGvogzxnyL9yjXbN+8u4H/9PXtKgJu8c2/B/irMeZneI+E3QHsbeE1HcB/+UKbAZ5poQ/ZRGCFMSYN+P7M3p6IdEfqQyYiAcHXh2yUtbbY37WIiJwuNVmKiIiI+JmOkImIiIj4mY6QiYiIiPiZApmIiIiInymQiYiIiPiZApmIiIiInymQiYiIiPjZ/wNirGpOs93xUwAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f884460d470>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# 绘图\n",
    "train_loss = history_dict['loss']\n",
    "val_loss = history_dict['val_loss']\n",
    "train_acc = history_dict['acc']\n",
    "val_acc = history_dict['val_acc']\n",
    "\n",
    "epochs = range(1, len(train_acc) + 1)\n",
    "\n",
    "plt.figure(figsize=(10, 8))\n",
    "\n",
    "plt.plot(epochs, train_loss, label='training_loss')\n",
    "plt.plot(epochs, val_loss, label='val_loss')\n",
    "plt.plot(epochs, train_acc, label='training_acc')\n",
    "plt.plot(epochs, val_acc, label='val_acc')\n",
    "plt.xlabel('Epochs #')\n",
    "plt.ylabel('Loss/Acc')\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 任务：\n",
    "有兴趣的同学，可以尝试下继续训练模型，更换参数，改变隐藏层的units数量，如32，64，128...<br>\n",
    "更换Adam/SGD等优化器\n",
    "更换mse loss function做对比"
   ]
  },
  {
   "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": "3.5.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}