1.2 Word Embeddings for IMDB.ipynb

{"nbformat":4,"nbformat_minor":0,"metadata":{"colab":{"name":"1.2 Word Embeddings for IMDB.ipynb","version":"0.3.2","provenance":[{"file_id":"1IwWqmUBkEwwUAt7B8Vqa1ebTuubBWBBx","timestamp":1540392609033}]},"kernelspec":{"display_name":"Python 3","language":"python","name":"python3"}},"cells":[{"metadata":{"id":"yMs23mdipZQY","colab_type":"code","colab":{}},"cell_type":"code","source":["#Based on https://github.com/fchollet/deep-learning-with-python-notebooks/blob/master/6.1-using-word-embeddings.ipynb"],"execution_count":0,"outputs":[]},{"metadata":{"id":"_TKEzjLWYY7-","colab_type":"code","outputId":"a4047b37-0d7d-4d14-d11d-b827af5dc83e","executionInfo":{"status":"ok","timestamp":1540395289978,"user_tz":-120,"elapsed":560,"user":{"displayName":"Krzysztof Suwada","photoUrl":"","userId":"18334398126529369335"}},"colab":{"base_uri":"https://localhost:8080/","height":34}},"cell_type":"code","source":["import tensorflow as tf\n","import tensorflow.keras\n","tf.keras.__version__"],"execution_count":28,"outputs":[{"output_type":"execute_result","data":{"text/plain":["'2.1.6-tf'"]},"metadata":{"tags":[]},"execution_count":28}]},{"metadata":{"id":"wr7rQYe4YY8L","colab_type":"code","colab":{}},"cell_type":"code","source":["from tensorflow.keras.layers import Embedding\n","\n","# The Embedding layer takes at least two arguments:\n","# the number of possible tokens, here 1000 (1 + maximum word index),\n","# and the dimensionality of the embeddings, here 64.\n","embedding_layer = Embedding(1000, 64)"],"execution_count":0,"outputs":[]},{"metadata":{"id":"QVThfRWpYY8Q","colab_type":"code","colab":{}},"cell_type":"code","source":["from tensorflow.keras.datasets import imdb\n","from tensorflow.keras import preprocessing\n","\n","# Number of words to consider as features\n","max_features = 10000\n","# Cut texts after this number of words \n","# (among top max_features most common words)\n","maxlen = 20\n","\n","# Load the data as lists of integers.\n","(x_train, y_train), (x_test, y_test) = imdb.load_data(num_words=max_features)\n","\n","# This turns our lists of integers\n","# into a 2D integer tensor of shape `(samples, maxlen)`\n","x_train = preprocessing.sequence.pad_sequences(x_train, maxlen=maxlen)\n","x_test = preprocessing.sequence.pad_sequences(x_test, maxlen=maxlen)"],"execution_count":0,"outputs":[]},{"metadata":{"id":"Tq7IFAk-gLCh","colab_type":"code","outputId":"88155fd0-bc2c-489b-e565-b3c93af28024","executionInfo":{"status":"ok","timestamp":1540395302654,"user_tz":-120,"elapsed":699,"user":{"displayName":"Krzysztof Suwada","photoUrl":"","userId":"18334398126529369335"}},"colab":{"base_uri":"https://localhost:8080/","height":50}},"cell_type":"code","source":["x_train[0]"],"execution_count":31,"outputs":[{"output_type":"execute_result","data":{"text/plain":["array([  65,   16,   38, 1334,   88,   12,   16,  283,    5,   16, 4472,\n","        113,  103,   32,   15,   16, 5345,   19,  178,   32], dtype=int32)"]},"metadata":{"tags":[]},"execution_count":31}]},{"metadata":{"id":"KZABqlF8YY8T","colab_type":"code","outputId":"85cb3bf1-3f91-4fda-ed89-d8164e8c736f","executionInfo":{"status":"ok","timestamp":1540395315364,"user_tz":-120,"elapsed":12404,"user":{"displayName":"Krzysztof Suwada","photoUrl":"","userId":"18334398126529369335"}},"colab":{"base_uri":"https://localhost:8080/","height":642}},"cell_type":"code","source":["from tensorflow.keras.models import Sequential\n","from tensorflow.keras.layers import Flatten, Dense\n","\n","model = Sequential()\n","# We specify the maximum input length to our Embedding layer\n","# so we can later flatten the embedded inputs\n","model.add(Embedding(10000, 8, input_length=maxlen))\n","# After the Embedding layer, \n","# our activations have shape `(samples, maxlen, 8)`.\n","\n","# We flatten the 3D tensor of embeddings \n","# into a 2D tensor of shape `(samples, maxlen * 8)`\n","model.add(Flatten())\n","\n","# We add the classifier on top\n","model.add(Dense(1, activation='sigmoid'))\n","model.compile(optimizer='rmsprop', loss='binary_crossentropy', metrics=['acc'])\n","model.summary()\n","\n","history = model.fit(x_train, y_train,\n","                    epochs=10,\n","                    batch_size=32,\n","                    validation_split=0.2)"],"execution_count":32,"outputs":[{"output_type":"stream","text":["_________________________________________________________________\n","Layer (type)                 Output Shape              Param #   \n","=================================================================\n","embedding_1 (Embedding)      (None, 20, 8)             80000     \n","_________________________________________________________________\n","flatten (Flatten)            (None, 160)               0         \n","_________________________________________________________________\n","dense (Dense)                (None, 1)                 161       \n","=================================================================\n","Total params: 80,161\n","Trainable params: 80,161\n","Non-trainable params: 0\n","_________________________________________________________________\n"],"name":"stdout"},{"output_type":"stream","text":["/usr/local/lib/python3.6/dist-packages/tensorflow/python/ops/gradients_impl.py:112: UserWarning: Converting sparse IndexedSlices to a dense Tensor of unknown shape. This may consume a large amount of memory.\n","  \"Converting sparse IndexedSlices to a dense Tensor of unknown shape. \"\n"],"name":"stderr"},{"output_type":"stream","text":["Train on 20000 samples, validate on 5000 samples\n","Epoch 1/10\n","20000/20000 [==============================] - 1s 71us/step - loss: 0.6598 - acc: 0.6411 - val_loss: 0.5997 - val_acc: 0.7092\n","Epoch 2/10\n","20000/20000 [==============================] - 1s 57us/step - loss: 0.5276 - acc: 0.7581 - val_loss: 0.5178 - val_acc: 0.7360\n","Epoch 3/10\n","20000/20000 [==============================] - 1s 55us/step - loss: 0.4569 - acc: 0.7900 - val_loss: 0.4970 - val_acc: 0.7502\n","Epoch 4/10\n","20000/20000 [==============================] - 1s 56us/step - loss: 0.4215 - acc: 0.8092 - val_loss: 0.4927 - val_acc: 0.7538\n","Epoch 5/10\n","20000/20000 [==============================] - 1s 56us/step - loss: 0.3968 - acc: 0.8219 - val_loss: 0.4928 - val_acc: 0.7572\n","Epoch 6/10\n","20000/20000 [==============================] - 1s 56us/step - loss: 0.3763 - acc: 0.8326 - val_loss: 0.4963 - val_acc: 0.7598\n","Epoch 7/10\n","20000/20000 [==============================] - 1s 56us/step - loss: 0.3582 - acc: 0.8439 - val_loss: 0.5033 - val_acc: 0.7564\n","Epoch 8/10\n","20000/20000 [==============================] - 1s 56us/step - loss: 0.3415 - acc: 0.8538 - val_loss: 0.5062 - val_acc: 0.7542\n","Epoch 9/10\n","20000/20000 [==============================] - 1s 56us/step - loss: 0.3251 - acc: 0.8636 - val_loss: 0.5125 - val_acc: 0.7528\n","Epoch 10/10\n","20000/20000 [==============================] - 1s 56us/step - loss: 0.3090 - acc: 0.8728 - val_loss: 0.5208 - val_acc: 0.7524\n"],"name":"stdout"}]},{"metadata":{"id":"FsZnW9-zZo0P","colab_type":"code","outputId":"405304f3-cb0a-405a-c4de-3afaf82483f0","executionInfo":{"status":"ok","timestamp":1540395325068,"user_tz":-120,"elapsed":9681,"user":{"displayName":"Krzysztof Suwada","photoUrl":"","userId":"18334398126529369335"}},"colab":{"base_uri":"https://localhost:8080/","height":67}},"cell_type":"code","source":["!rm -rf aclImdb_v1.tar.gz*\n","!curl -O http://ai.stanford.edu/~amaas/data/sentiment/aclImdb_v1.tar.gz"],"execution_count":33,"outputs":[{"output_type":"stream","text":["  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current\n","                                 Dload  Upload   Total   Spent    Left  Speed\n","100 80.2M  100 80.2M    0     0  13.3M      0  0:00:05  0:00:05 --:--:-- 17.7M\n"],"name":"stdout"}]},{"metadata":{"id":"OwaBD_ywZuw4","colab_type":"code","colab":{}},"cell_type":"code","source":["!tar -zxf aclImdb_v1.tar.gz "],"execution_count":0,"outputs":[]},{"metadata":{"id":"fEIeBN8KYY8a","colab_type":"code","colab":{}},"cell_type":"code","source":["import os\n","\n","imdb_dir = './aclImdb'\n","train_dir = os.path.join(imdb_dir, 'train')\n","\n","labels = []\n","texts = []\n","\n","for label_type in ['neg', 'pos']:\n","    dir_name = os.path.join(train_dir, label_type)\n","    for fname in os.listdir(dir_name):\n","        if fname[-4:] == '.txt':\n","            f = open(os.path.join(dir_name, fname))\n","            texts.append(f.read())\n","            f.close()\n","            if label_type == 'neg':\n","                labels.append(0)\n","            else:\n","                labels.append(1)"],"execution_count":0,"outputs":[]},{"metadata":{"id":"0QMlYY5JYY8d","colab_type":"text"},"cell_type":"markdown","source":["### Tokenize the data\n","\n","\n","Let's vectorize the texts we collected, and prepare a training and validation split.\n","We will merely be using the concepts we introduced earlier in this section.\n","\n","Because pre-trained word embeddings are meant to be particularly useful on problems where little training data is available (otherwise, \n","task-specific embeddings are likely to outperform them), we will add the following twist: we restrict the training data to its first 200 \n","samples. So we will be learning to classify movie reviews after looking at just 200 examples...\n"]},{"metadata":{"id":"y8QcAkTVYY8e","colab_type":"code","outputId":"08613ba7-078e-436e-f68b-37b311db0277","executionInfo":{"status":"ok","timestamp":1540395354183,"user_tz":-120,"elapsed":12052,"user":{"displayName":"Krzysztof Suwada","photoUrl":"","userId":"18334398126529369335"}},"colab":{"base_uri":"https://localhost:8080/","height":67}},"cell_type":"code","source":["from tensorflow.keras.preprocessing.text import Tokenizer\n","from tensorflow.keras.preprocessing.sequence import pad_sequences\n","import numpy as np\n","\n","maxlen = 100  # We will cut reviews after 100 words\n","training_samples = 20000  # We will be training on 200 samples\n","validation_samples = 10000  # We will be validating on 10000 samples\n","max_words = 10000  # We will only consider the top 10,000 words in the dataset\n","\n","tokenizer = Tokenizer(num_words=max_words)\n","tokenizer.fit_on_texts(texts)\n","sequences = tokenizer.texts_to_sequences(texts)\n","\n","word_index = tokenizer.word_index\n","print('Found %s unique tokens.' % len(word_index))\n","\n","data = pad_sequences(sequences, maxlen=maxlen)\n","\n","labels = np.asarray(labels)\n","print('Shape of data tensor:', data.shape)\n","print('Shape of label tensor:', labels.shape)\n","\n","# Split the data into a training set and a validation set\n","# But first, shuffle the data, since we started from data\n","# where sample are ordered (all negative first, then all positive).\n","indices = np.arange(data.shape[0])\n","np.random.shuffle(indices)\n","data = data[indices]\n","labels = labels[indices]\n","\n","x_train = data[:training_samples]\n","y_train = labels[:training_samples]\n","x_val = data[training_samples: training_samples + validation_samples]\n","y_val = labels[training_samples: training_samples + validation_samples]"],"execution_count":36,"outputs":[{"output_type":"stream","text":["Found 88582 unique tokens.\n","Shape of data tensor: (25000, 100)\n","Shape of label tensor: (25000,)\n"],"name":"stdout"}]},{"metadata":{"id":"CNj9Cb8oa7U_","colab_type":"code","outputId":"3ef2503e-5f96-481a-c855-a0007f8d026f","executionInfo":{"status":"ok","timestamp":1540395355566,"user_tz":-120,"elapsed":539,"user":{"displayName":"Krzysztof Suwada","photoUrl":"","userId":"18334398126529369335"}},"colab":{"base_uri":"https://localhost:8080/","height":185}},"cell_type":"code","source":["x_train[0]"],"execution_count":37,"outputs":[{"output_type":"execute_result","data":{"text/plain":["array([  10,   13, 2231,   58, 1748,    6,    8,   46, 3832,   26,   13,\n","          8,    3, 2549, 4102, 1694,  294, 1712, 2149,   91,  324,  316,\n","         87,   20,  245,   26,   13,   79,   20, 1459, 4140, 1374,  546,\n","         26,  207,   76,    1,  389,  170, 1078, 1078,  135,   78,   72,\n","         25,    5,  205,  155,  408, 1581,   35,  376,   10,  101,  143,\n","        221,  112,  327,  143,  128,   21,  221,   48,    6,   11,    3,\n","        973,  135,   78,   72,   25,    5,  137,   29,    1,   93,    5,\n","        344,  744,   63,  802,    1,  210,    4,    9,   10,   77,   40,\n","        205, 1522,  712,   15,  147,   62,    6,  250,   10,  116,   58,\n","        365], dtype=int32)"]},"metadata":{"tags":[]},"execution_count":37}]},{"metadata":{"id":"CGGv6zXRYY8i","colab_type":"text"},"cell_type":"markdown","source":["### Download the GloVe word embeddings\n","\n","\n","Head to `https://nlp.stanford.edu/projects/glove/` (where you can learn more about the GloVe algorithm), and download the pre-computed \n","embeddings from 2014 English Wikipedia. It's a 822MB zip file named `glove.6B.zip`, containing 100-dimensional embedding vectors for \n","400,000 words (or non-word tokens). Un-zip it."]},{"metadata":{"id":"28WWZsIgYY8j","colab_type":"text"},"cell_type":"markdown","source":["### Pre-process the embeddings\n","\n","\n","Let's parse the un-zipped file (it's a `txt` file) to build an index mapping words (as strings) to their vector representation (as number \n","vectors)."]},{"metadata":{"id":"A0-99aUabJNR","colab_type":"code","outputId":"232275df-00d9-46d1-91c7-811c12504687","executionInfo":{"status":"ok","timestamp":1540395719464,"user_tz":-120,"elapsed":363491,"user":{"displayName":"Krzysztof Suwada","photoUrl":"","userId":"18334398126529369335"}},"colab":{"base_uri":"https://localhost:8080/","height":67}},"cell_type":"code","source":["!rm -rf glove.6B.zip*\n","!curl -O https://nlp.stanford.edu/data/glove.6B.zip"],"execution_count":38,"outputs":[{"output_type":"stream","text":["  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current\n","                                 Dload  Upload   Total   Spent    Left  Speed\n","100  822M  100  822M    0     0  2339k      0  0:05:59  0:05:59 --:--:-- 4849k\n"],"name":"stdout"}]},{"metadata":{"id":"lgfrl0guohXg","colab_type":"code","colab":{"base_uri":"https://localhost:8080/","height":286},"outputId":"d8f4b60b-5282-46c7-e419-35b7550cc198","executionInfo":{"status":"ok","timestamp":1540395722024,"user_tz":-120,"elapsed":2510,"user":{"displayName":"Krzysztof Suwada","photoUrl":"","userId":"18334398126529369335"}}},"cell_type":"code","source":["!ls -lah"],"execution_count":39,"outputs":[{"output_type":"stream","text":["total 3.0G\n","drwxr-xr-x 1 root root 4.0K Oct 24 15:36 .\n","drwxr-xr-x 1 root root 4.0K Oct 24 11:44 ..\n","drwxr-xr-x 4 7297 1000 4.0K Jun 26  2011 aclImdb\n","-rw-r--r-- 1 root root  81M Oct 24 15:35 aclImdb_v1.tar.gz\n","drwxr-xr-x 4 root root 4.0K Oct 22 18:12 .config\n","drwxr-xr-x 2 root root 4.0K Oct 24 12:00 gdline_frames\n","-rw-r--r-- 1 root root 6.3K Oct 24 12:00 gdline.htm\n","-rw-rw-r-- 1 root root 332M Aug  4  2014 glove.6B.100d.txt\n","-rw-rw-r-- 1 root root 662M Aug  4  2014 glove.6B.200d.txt\n","-rw-rw-r-- 1 root root 990M Aug 27  2014 glove.6B.300d.txt\n","-rw-rw-r-- 1 root root 164M Aug  4  2014 glove.6B.50d.txt\n","-rw-r--r-- 1 root root 823M Oct 24 15:42 glove.6B.zip\n","-rw-r--r-- 1 root root    0 Oct 24 12:00 None0000000.png\n","-rw-r--r-- 1 root root 6.3M Oct 24 15:16 pre_trained_glove_model.h5\n","drwxr-xr-x 2 root root 4.0K Oct 22 18:24 sample_data\n"],"name":"stdout"}]},{"metadata":{"id":"p45tsqGgbMBd","colab_type":"code","outputId":"ac079bf6-614f-4dc6-a5a6-4f38c210e18c","executionInfo":{"status":"ok","timestamp":1539264159527,"user_tz":-120,"elapsed":50326,"user":{"displayName":"Krzysztof Suwada","photoUrl":"","userId":"18334398126529369335"}},"colab":{"base_uri":"https://localhost:8080/","height":78}},"cell_type":"code","source":["!unzip glove.6B.zip"],"execution_count":0,"outputs":[{"output_type":"stream","text":["Archive:  glove.6B.zip\n","replace glove.6B.50d.txt? [y]es, [n]o, [A]ll, [N]one, [r]ename: "],"name":"stdout"}]},{"metadata":{"id":"xligdB-OYY8j","colab_type":"code","outputId":"37b08587-13fb-4674-9d0d-040c15a8a5ef","executionInfo":{"status":"ok","timestamp":1540394040158,"user_tz":-120,"elapsed":13018,"user":{"displayName":"Krzysztof Suwada","photoUrl":"","userId":"18334398126529369335"}},"colab":{"base_uri":"https://localhost:8080/","height":34}},"cell_type":"code","source":["glove_dir = './'\n","\n","\n","embeddings_index = {}\n","f = open(os.path.join(glove_dir, 'glove.6B.100d.txt'))\n","for line in f:\n","    values = line.split()\n","    word = values[0]\n","    coefs = np.asarray(values[1:], dtype='float32')\n","    embeddings_index[word] = coefs\n","f.close()\n","\n","print('Found %s word vectors.' % len(embeddings_index))"],"execution_count":20,"outputs":[{"output_type":"stream","text":["Found 400000 word vectors.\n"],"name":"stdout"}]},{"metadata":{"id":"ZEZrhUMxYY8n","colab_type":"code","colab":{}},"cell_type":"code","source":["embedding_dim = 100\n","\n","embedding_matrix = np.zeros((max_words, embedding_dim))\n","for word, i in word_index.items():\n","    embedding_vector = embeddings_index.get(word)\n","    if i < max_words:\n","        if embedding_vector is not None:\n","            # Words not found in embedding index will be all-zeros.\n","            embedding_matrix[i] = embedding_vector"],"execution_count":0,"outputs":[]},{"metadata":{"id":"MteY1X9NYY8r","colab_type":"code","outputId":"9f34365b-9417-4d5f-9b7c-734bbb80dd1a","executionInfo":{"status":"ok","timestamp":1540394105224,"user_tz":-120,"elapsed":583,"user":{"displayName":"Krzysztof Suwada","photoUrl":"","userId":"18334398126529369335"}},"colab":{"base_uri":"https://localhost:8080/","height":386}},"cell_type":"code","source":["from keras.models import Sequential\n","from keras.layers import Embedding, Flatten, Dense, Dropout\n","\n","model = Sequential()\n","model.add(Embedding(max_words, embedding_dim, input_length=maxlen))\n","model.add(Flatten())\n","model.add(Dense(64, activation='relu'))\n","model.add(Dropout(0.5))\n","model.add(Dense(32, activation='relu'))\n","model.add(Dropout(0.5))\n","model.add(Dense(1, activation='sigmoid'))\n","model.summary()"],"execution_count":22,"outputs":[{"output_type":"stream","text":["_________________________________________________________________\n","Layer (type)                 Output Shape              Param #   \n","=================================================================\n","embedding_1 (Embedding)      (None, 100, 100)          1000000   \n","_________________________________________________________________\n","flatten_1 (Flatten)          (None, 10000)             0         \n","_________________________________________________________________\n","dense_1 (Dense)              (None, 64)                640064    \n","_________________________________________________________________\n","dropout_1 (Dropout)          (None, 64)                0         \n","_________________________________________________________________\n","dense_2 (Dense)              (None, 32)                2080      \n","_________________________________________________________________\n","dropout_2 (Dropout)          (None, 32)                0         \n","_________________________________________________________________\n","dense_3 (Dense)              (None, 1)                 33        \n","=================================================================\n","Total params: 1,642,177\n","Trainable params: 1,642,177\n","Non-trainable params: 0\n","_________________________________________________________________\n"],"name":"stdout"},{"output_type":"stream","text":["Using TensorFlow backend.\n"],"name":"stderr"}]},{"metadata":{"id":"xZnD5Qu7YY8u","colab_type":"code","colab":{}},"cell_type":"code","source":["model.layers[0].set_weights([embedding_matrix])\n","model.layers[0].trainable = False"],"execution_count":0,"outputs":[]},{"metadata":{"id":"Jh5geMNuYY82","colab_type":"code","outputId":"c0967291-8bcf-4152-f073-e8936ce04749","executionInfo":{"status":"ok","timestamp":1540394196832,"user_tz":-120,"elapsed":88537,"user":{"displayName":"Krzysztof Suwada","photoUrl":"","userId":"18334398126529369335"}},"colab":{"base_uri":"https://localhost:8080/","height":370}},"cell_type":"code","source":["model.compile(optimizer='rmsprop',\n","              loss='binary_crossentropy',\n","              metrics=['acc'])\n","history = model.fit(x_train, y_train,\n","                    epochs=10,\n","                    batch_size=32,\n","                    validation_data=(x_val, y_val))\n","model.save_weights('pre_trained_glove_model.h5')"],"execution_count":24,"outputs":[{"output_type":"stream","text":["Train on 20000 samples, validate on 5000 samples\n","Epoch 1/10\n","20000/20000 [==============================] - 9s 456us/step - loss: 0.7172 - acc: 0.4998 - val_loss: 0.6931 - val_acc: 0.5028\n","Epoch 2/10\n","20000/20000 [==============================] - 9s 435us/step - loss: 0.6964 - acc: 0.5037 - val_loss: 0.6917 - val_acc: 0.5230\n","Epoch 3/10\n","20000/20000 [==============================] - 9s 432us/step - loss: 0.6900 - acc: 0.5422 - val_loss: 0.6525 - val_acc: 0.6472\n","Epoch 4/10\n","20000/20000 [==============================] - 9s 434us/step - loss: 0.6714 - acc: 0.5797 - val_loss: 0.6525 - val_acc: 0.6306\n","Epoch 5/10\n","20000/20000 [==============================] - 9s 442us/step - loss: 0.6503 - acc: 0.6155 - val_loss: 0.6398 - val_acc: 0.6304\n","Epoch 6/10\n","20000/20000 [==============================] - 9s 440us/step - loss: 0.6278 - acc: 0.6434 - val_loss: 0.5952 - val_acc: 0.7154\n","Epoch 7/10\n","20000/20000 [==============================] - 9s 440us/step - loss: 0.6039 - acc: 0.6777 - val_loss: 0.5765 - val_acc: 0.7210\n","Epoch 8/10\n","20000/20000 [==============================] - 9s 435us/step - loss: 0.5841 - acc: 0.6948 - val_loss: 0.5867 - val_acc: 0.6996\n","Epoch 9/10\n","20000/20000 [==============================] - 8s 423us/step - loss: 0.5589 - acc: 0.7185 - val_loss: 0.6096 - val_acc: 0.7054\n","Epoch 10/10\n","20000/20000 [==============================] - 9s 433us/step - loss: 0.5389 - acc: 0.7357 - val_loss: 0.5881 - val_acc: 0.6890\n"],"name":"stdout"}]},{"metadata":{"id":"kXUYJ0sMYY85","colab_type":"text"},"cell_type":"markdown","source":["Let's plot its performance over time:"]},{"metadata":{"id":"syk4JOKqYY86","colab_type":"code","outputId":"daa76fb9-2290-4621-df65-b2a1211919d1","executionInfo":{"status":"ok","timestamp":1540394251257,"user_tz":-120,"elapsed":965,"user":{"displayName":"Krzysztof Suwada","photoUrl":"","userId":"18334398126529369335"}},"colab":{"base_uri":"https://localhost:8080/","height":706}},"cell_type":"code","source":["import matplotlib.pyplot as plt\n","\n","acc = history.history['acc']\n","val_acc = history.history['val_acc']\n","loss = history.history['loss']\n","val_loss = history.history['val_loss']\n","\n","epochs = range(1, len(acc) + 1)\n","\n","plt.plot(epochs, acc, 'bo', label='Training acc')\n","plt.plot(epochs, val_acc, 'b', label='Validation acc')\n","plt.title('Training and validation accuracy')\n","plt.legend()\n","\n","plt.figure()\n","\n","plt.plot(epochs, loss, 'bo', label='Training loss')\n","plt.plot(epochs, val_loss, 'b', label='Validation loss')\n","plt.title('Training and validation loss')\n","plt.legend()\n","\n","plt.show()"],"execution_count":25,"outputs":[{"output_type":"display_data","data":{"image/png":"iVBORw0KGgoAAAANSUhEUgAAAecAAAFZCAYAAACizedRAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n","text/plain":["<matplotlib.figure.Figure at 0x7f53e47c8198>"]},"metadata":{"tags":[]}},{"output_type":"display_data","data":{"image/png":"iVBORw0KGgoAAAANSUhEUgAAAe0AAAFZCAYAAAC173eYAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMS4yLCBo\ndHRwOi8vbWF0cGxvdGxpYi5vcmcvNQv5yAAAIABJREFUeJzs3X98zXX/x/HH2Tn7dWZqY8NcdOEy\nGqlLXVchv2ZjSKwiFAlRkQmFhZEYMaGSVXIlql1clhQWiX7topCr/Ij09aOIze/93jk73z9OnVo2\nG7Zzdrbn/Xbr1j7v8/nx/rydc17n83q/P++PwWaz2RAREZEKz8PVFRAREZHSUdAWERFxEwraIiIi\nbkJBW0RExE0oaIuIiLgJBW0RERE3oaAtVVJcXBxRUVFERUXRrFkzOnbs6FjOyMi4on1FRUWRnp5+\n2XUSEhJ45513rqXKZW7QoEGsXr26TPbVpEkTfvnlFzZu3MjEiROv6Xj//ve/HX+Xpm1La8KECSxa\ntKhM9iXiKiZXV0DEFaZNm+b4Ozw8nOeff57bbrvtqva1YcOGEtcZO3bsVe3b3URGRhIZGXnV26el\npfH666/Tp08foHRtK1KV6EpbpAgDBgzghRdeoGvXruzcuZP09HSGDBlCVFQU4eHhLF261LHub1eZ\n27Zt4/777ychIYGuXbsSHh7O9u3bgcJXeeHh4bz77rvcd9993HnnncyaNcuxr8WLF9OqVSvuvfde\nVqxYQXh4eJH1W7lyJV27dqVz58488MAD/PzzzwCsXr2aUaNGERsbS5cuXejWrRsHDx4E4NixY/Tu\n3ZuIiAjGjh2L1Wq9ZL9bt26lR48ehcp69uzJp59+etk2+M3q1asZNGhQicf7+OOP6dGjB126dOGe\ne+5h3759APTt25fjx48TFRVFXl6eo20Bli1bRrdu3YiKiuKxxx7jzJkzjrZduHAhDz/8MB07duTh\nhx8mOzu7uH9aAPbv30/fvn2JioqiZ8+efPbZZwBkZmYyYsQIunbtSqdOnZg0aRL5+fnFlos4m4K2\nSDG+++47PvzwQ1q2bMkrr7zCX/7yFzZs2MCbb75JQkICJ06cuGSbvXv3cvPNN7N+/Xr69+/PK6+8\nUuS+v/rqK5KSkvjPf/7D8uXL+eWXXzh48CCvv/46a9as4e233y72KvP06dM8++yzLF26lI8++oj6\n9esXSvt++umn9O/fn5SUFG6//XbefPNNAObOnUurVq3YtGkTDz30EDt37rxk361ateKXX37h2LFj\ngD3w/vLLL7Ru3brUbfCb4o5nsViYMGEC06dPJyUlhfDwcGbPng3AzJkzqVOnDhs2bMDLy8uxr2++\n+YYlS5bw1ltvsWHDBkJCQkhISHC8vmHDBl544QU2btzImTNn2LhxY7H1KigoYMyYMTz44INs2LCB\n5557jrFjx5KRkcF7771H9erVWb9+PSkpKRiNRn744Ydiy0WcTUFbpBjt27fHw8P+EZk0aRKTJ08G\noF69egQFBfHTTz9dso2fnx8REREANGvWjOPHjxe57x49emA0GqlVqxY1atTgxIkTfPXVV/zzn/8k\nODgYb29v7r333iK3rVGjBjt27KB27doA3HbbbY4gC9CoUSOaN28OQFhYmCOwfv3113Tr1g2AFi1a\n0LBhw0v27eXlRceOHdm8eTMAmzZtIiIiApPJVOo2+E1xxzOZTHz55ZfccsstRda/KFu2bKFLly7U\nqFEDgN69e/PFF184Xm/fvj3XX389JpOJ0NDQy/6Y+Omnn0hPT6d79+4A3HTTTYSEhPDtt98SGBjI\nrl27+PzzzykoKGDatGnceOONxZaLOJv6tEWKcd111zn+/vbbbx1Xlh4eHqSlpVFQUHDJNv7+/o6/\nPTw8ilwHoFq1ao6/jUYjVquVCxcuFDpmrVq1itzWarWycOFCNm/ejNVqJTMzkwYNGhRZh9/2DXD+\n/PlCx61evXqR++/SpQvLli3joYceYtOmTTz++ONX1Aa/udzx3nrrLZKTk8nLyyMvLw+DwVDsfgDO\nnDlDcHBwoX2dPn26xHMubl/+/v6Fjlm9enXOnDlD9+7dOX/+PAsWLODHH3/k7rvvZuLEiXTt2rXI\n8j9mA0ScQVfaIqXw1FNP0aVLF1JSUtiwYQMBAQFlfoxq1aqRlZXlWD516lSR661bt47NmzezfPly\nUlJSGDVqVKn2X7169UIj43/rE/6ztm3bsn//fg4fPszhw4e54447gCtvg+KOt3PnTl577TVeeeUV\nUlJSeO6550qse82aNTl37pxj+dy5c9SsWbPE7YpSo0YNzp8/zx+flXTu3DnHVXzfvn1ZuXIl69at\nY8+ePbz33nuXLRdxJgVtkVI4ffo0zZs3x2AwkJycTHZ2dqEAWxZatGjBtm3bOHPmDHl5ecUGhdOn\nT1O3bl0CAwM5e/Ys69evJzMzs8T933LLLY6+3p07d3L06NEi1/Py8uLOO+9kzpw5dOrUCaPR6Dju\nlbRBccc7c+YMNWrUICQkhOzsbJKTk8nKysJms2EymcjKysJisRTaV4cOHdi4cSNnz54F4N1336V9\n+/YlnnNR/vKXv1C7dm3WrVvnqFt6ejotWrTg5ZdfZtWqVYA90/GXv/wFg8FQbLmIsyloi5RCTEwM\nI0aMoEePHmRlZXH//fczefLkYgPf1WjRogXR0dFER0czcOBAOnbsWOR6d911F+fOnSMyMpKxY8cy\nevRofvnll0Kj0Ivy1FNP8cknnxAREcGKFSto3bp1set26dKFTZs20bVrV0fZlbZBccdr27YtwcHB\nREREMHjwYB566CH8/f0ZNWoUTZo04brrrqNNmzaFxgO0aNGCYcOG8cADDxAVFcXFixd58sknL3u+\nxTEYDMybN4/ly5fTtWtXnnvuORYsWIDZbKZnz56sWbOGLl26EBUVhaenJz179iy2XMTZDHqetkjF\nYbPZHFdwW7ZsYf78+UrDioiDrrRFKogzZ85wxx138PPPP2Oz2Vi/fr1jhLWICOhKW6RCeeedd3jj\njTcwGAw0bNiQGTNmOAZIiYgoaIuIiLiJUt2nPXPmTHbv3o3BYCA2NpYWLVoAcPLkScaNG+dY79ix\nY4wdO5auXbvyzDPPcPToUaxWK08//TS33XYbAwYMICsrC7PZDMD48eMdk0CIiIjI5ZUYtLdv386R\nI0dISkri0KFDxMbGkpSUBNhvfXjrrbcA+9SEAwYMIDw8nDVr1uDr68s777zDwYMHmThxouN2ifj4\neEJDQ8vxlERERCqnEoN2amqqY1rGRo0acf78eTIyMgrNdASQnJxMly5d8PPz4+677+auu+4CIDAw\nsNCkCFciLe3iVW1X2QQEmDl7tmzvCZZLqZ2dQ+3sHGpn5ynrtg4K8i/2tRKDdnp6Os2aNXMsBwYG\nkpaWdknQXrlyJW+88QYAnp6ejvI333zTEcABFi5cyNmzZ2nUqBGxsbH4+PgUe+yAADMmk7GkKlYJ\nl/tHlLKjdnYOtbNzqJ2dx1ltfcVzjxc1bm3Xrl00bNjwkkC+YsUK9uzZw+LFiwEYOHAgTZo0oX79\n+sTFxbFixQqGDBlS7LH0K9EuKMhfWQcnUDs7h9rZOdTOzlPWbX25HwAl3qcdHBxMenq6Y/nUqVME\nBQUVWmfLli20atWqUNnKlSvZvHkzixYtclx5R0ZGUr9+fcD+TOEDBw6U/ixERESquBKDdps2bUhJ\nSQFgz549BAcHX3JF/e2339K0aVPH8rFjx3j33Xd56aWX8Pb2BuxX6IMGDeLChQsAbNu2jcaNG5fZ\niYiIiFR2JabHW7ZsSbNmzejbty8Gg4G4uDhWr16Nv78/kZGRAKSlpRWaAGLlypWcO3eOYcOGOcqW\nLFlCnz59GDRoEL6+vtSqVYsnnniiHE5JRESkcqrQk6uoP8ZOfVPOoXZ2DrWzc6idnadC9WmLiIhI\nxaCgLSIi4iau+JYvERGRq/Xiiy/w/ff7OHPmNDk5OYSE1KV69euYOXNOiduuW7cWP79qtG9f9LPm\nFyxIoHfvvoSE1L2quo0cOYwxY56mYcO/XdX2zlBlgnZyson58704cMCD0NACRo/OIzra4upqiYhU\naGX93fnEE08C9gD844+HGDlydKm37datx2Vfj4kZe9X1chdVImgnJ5sYPtzXsbxvn/HX5WwFbhGR\nYjjzu3Pnzq95993lZGVlMXLkk+zatYMtWz6moKCAVq3aMHjwMJYsSeT666+nQYNGrF79bwwGD44c\n+T86dOjE4MHDHFfKn3zyMZmZGRw9eoSff/6JUaPG0qpVG5Yv/xebNn1ESEhdLBYLffs+QMuWt11S\nl4yMDGbMmEpGxkUsFgujRz9FkyZNmT9/Dvv378NqtRIdfR/duvVg/vw5HDp0gJycPEdZeaoSQXv+\nfK8iyxcs8FLQFhEphrO/Ow8d+oF33lmNl5cXu3btYNGi1/Hw8KBPn57cf3//Quvu3buHt9/+DwUF\nBfTu3YPBg4cVev3UqZPMnbuQ//73S9as+Q/NmjVn9eqVvPPOf8jMzKRv33vo2/eBIuuxcuU7NGvW\nnAcfHMT+/Xt58cV5zJw5hy+//Jx//3sNFouFdevWcuHCeb788nM++WQzJ06cZd26tWXeJn9WJYL2\ngQNFj7crrlxERJz/3fm3vzXGy8v+Q8HHx4eRI4dhNBo5d+6cY2Ku3zRp0vSyz65o0eIWwD6rZ0ZG\nBj/9dIyGDRvh7e2Dt7cPN97YrNht9+/fy8CB9im2mzYN46efjlG9+nXUq3cDEyaMoWPHCKKiuuPl\n5UW9ejfw2GOP0aZNB6Kiul9rE5SoSkSt0NCCKyoXERHnf3f+NuX1L7+cIClpBQkJL/LSS69Su3bt\nS9Y1Gi//MKk/vm6z2bDZwMPj95BnMBS/rcFgKPScjYIC+/kmJCzk4YeHcfDgAcaPf9JRNnLkyEJl\n5alKBO3Ro/OKLI+JKbpcRERc99157tw5AgICMJvNfP/9fn755Rfy8/OvaZ916tThxx8PYbFYOHv2\nLPv37yt23aZNw9i162sAvvvuWxo0aMSJE8dZufJdmjRpysiRozl//ryjrFmzZo6y8lYl0uP2vpds\nFiz4fQRkTIxGj4uIXI6rvjsbNw7F19fMY48N5qabbqFnz3tISJhNixY3X/U+AwNrEBkZxSOPDOSG\nGxoQFtas2Kv1Pn36MXPmNEaNepSCggLGjBlPzZpBfPfdbj7++CM8PT3p3v1uR1nfvn0BD7p3v/uq\n61damsbUDWg6QudQOzuH2tk51M6XWrduLZGRURiNRgYO7Mu8eS8SHFzrmvfrzGlMq8SVtoiIyOnT\npxk27CE8Pb3o3DmqTAK2syloi4hIlTBgwCAGDBjk6mpckyoxEE1ERKQyUNAWERFxEwraIiIibkJB\nW0RExE0oaIuIiNMMH/7wJRObLF78Eu+8s7zI9Xfu/JpJk54GYMKEMZe8/p//JLFkSWKxx/vhh4Mc\nPXoEgLi4ieTm5lxt1bnvvh5kZWVd9fZlQUFbREScJjKyC5s3byxUtmXLZiIiOpe47axZ8674eFu3\nbubYsaMATJsWj7d38fOVuwPd8iUiIk7TqVNnHntsCI8/PgqA/fv3ERQURFBQMF99tY3XX1+Mp6cn\n/v7+PPvsrELbdu/eiQ8//Jivv97OwoUJBAbWoEaNmo5Hbc6YMZW0tFNkZ2czePAwateuw5o1q9m6\ndTMBAQFMmTKRZcuSyMi4SHz8s+Tn5+Ph4cGECZMxGAzMmDGVkJC6/PDDQUJDmzBhwuQiz+HUqZOF\ntn/++VmYTNV49tnJnD6dTl5eHkOGDOe22/55Sdkdd7S+pvZT0BYRqaKmTvVm7dqyDQM9eliYOjW3\n2NcDAgIJCanL3r3fERbWnM2bNxIZGQXAxYsXiYt7jpCQukyfPoVt21Ixm82X7CMx8SUmT55O48ah\njBs3ipCQuly8eIF//vMOuna9i59//onJkyfwxhvLuf32VnTo0ImwsOaO7V9/fTF33dWTTp0688kn\nm3jjjVcZMmQ433+/j2nTZhIQEEh0dDcuXryIv/+ls5P9efuXXnqJHj3u4/z5c7z88mtcvHiR1NQv\nOHToh0vKrpXS4yIi4lSRkVF8/LE9Rf7FF5/SoUMnAK6//npmz36OkSOHsWvXDi5cKPoBHCdOnKBx\n41AAbrmlJQD+/tXZt28Pjz02mBkzpha7LcD33+/j73+/FYCWLW/j4MHvAahbtx41atTEw8ODmjWD\nyMzMKNX2e/fu5YYb/kpWVibTp09m586viIjoXGTZtdKVtohIFTV1au5lr4rLS/v2HVm27A0iI7tQ\nr159qlevDkB8/HTmzJnPX//agHnzZhe7/R8fsfnb4zM2btzAhQsXePnl17lw4QJDhw64TA1+f/Rm\nfr4Fg8G+vz8/QKT4R3MU3t7DwwMfHx8SE//Ft9/+j/Xr1/LFF58RGxtXZNm10JW2iIg4ldnsR6NG\njVm2bKkjNQ6QmZlBrVq1uXjxIjt37ij2cZw1awZx9OhhbDYbu3btAOyP86xTJwQPDw+2bt3s2NZg\nMGC1Wgttf+ONYezcaX/05jff7KBp0xuvqP5/3r558+Z8//1+Nm7cwM0338K4cRM5fPj/iiy7VrrS\nFhERp4uMjOK55+KIi5vuKLvnnt489tgQ6tWrzwMPDOSNN15l2LDHL9l22LDHmTRpPLVr13E89KND\nh3AmTBjD3r3f0b373QQHB7N06WvcfPPfmT9/TqG+8aFDHyU+fjpr176HyeTJxImTsVhK/7jRP28/\nd+5sMjIsJCa+zJo1q/Hw8KB//wHUqRNySdm10qM53YAeseccamfnUDs7h9rZeZz5aE6lx0VERNyE\ngraIiIibUNAWERFxE6UaiDZz5kx2796NwWAgNjaWFi1aAHDy5EnGjRvnWO/YsWOMHTuWqKgoJkyY\nwPHjxzEajcTHx1OvXj3279/P1KlTAWjSpAnTpk0r+zMSERGppEq80t6+fTtHjhwhKSmJGTNmMGPG\nDMdrtWrV4q233uKtt95i6dKl1KlTh/DwcD744AOqV6/OO++8w6OPPkpCQgIAM2bMIDY2lnfffZeM\njAy2bt1afmcmIiJSyZQYtFNTU4mIiACgUaNGnD9/noyMS2eJSU5OpkuXLvj5+ZGamkpkZCQArVu3\nZufOneTl5fHzzz87rtI7duxIampqWZ6LiIhIpVZi0E5PTycgIMCxHBgYSFpa2iXrrVy5kvvuu8+x\nTWBgoP0AHh4YDAbS09Mds94A1KhRo8j9iIiISNGueHKVom7r3rVrFw0bNqRatWql3qY0t4cHBJgx\nmYwlrlcVXO6+PSk7amfnUDs7h9rZeZzV1iUG7eDgYNLT0x3Lp06dIigoqNA6W7ZsoVWrVoW2SUtL\no2nTpuTn52Oz2QgKCuLcuXOOdU6ePElwcPBlj332rGsfNl5RaJIE51A7O4fa2TnUzs5ToSZXadOm\nDSkpKQDs2bOH4ODgS66ov/32W5o2bVpomw0bNgDwySefcPvtt+Pp6UnDhg35+mv7fK0fffQRbdu2\nvfKzqUKSk020b2/GZIL27c0kJ2vWWRGRqqzEKNCyZUuaNWtG3759MRgMxMXFsXr1avz9/R2DzdLS\n0qhRo4Zjm27duvHll1/Sr18/vLy8mDXL/iDz2NhYpkyZQkFBATfffDOtW1/bw8Ars+RkE8OH+zqW\n9+0z/rqcTXR06efIFRGRykNzj1dQ7dub2bfv0v78sDArW7ao26A8KJ3oHGpn51A7O0+FSo+Laxw4\nUPQ/TXHlIiJS+SkCVFChoQVXVC4iIpWfgnYFNXp0XpHlMTFFl4uISOWnoF1BRUdbSEzMJizMislk\n78tOTNQgNBGRqqzK3ENkscAjj/hw8qQHfn42qlWz4ecHfn62X//jMmW/l5vNYHJSq0VHW4iOtvw6\nyEGDz0REqroqFbR//NGDgwc9sFgM17QvX1+bI4AXF9ztPwy4TNkfy8Goid9ERKQEVSZo+/jA1q32\nq9XcXMjMhMxMA5mZBjIyfv/7t/Kiygr/DRkZBo4f9yAjA6zW8vkhUK2ajZEjoWXLsmgFERFxZ1Um\naP+Rt7f9v8BAG3Dtt6nbbJCXhyOQFw7+hQP9b2VZWUX9aLD//88/BDZtgvXrPWjWrOKPHE9ONjF/\nvhcHDngQGlrA6NF56ocXESkjVTJolzWDoXx+COTmwsaNJoYM8WXoUF82bsykmGeyVAiaxU1EpHxp\n9HgFZTDYU/o9elgYMwYOHfJg3DgfKu78dTB/vleR5QsWFF0uIiJXRkHbDcyaBbfeamX1ak+WLfN0\ndXWKpVncRETKl75N3YCnJ7z2WjYBATYmTfLm228r5j+bZnETESlfFfPbXy7xl7/YePnlbHJzDQwZ\n4suFC66u0aU0i5uISPlS0HYjERFWnngil8OHPXjyyYrXv114FjebZnETESljGj3uZiZOzGP7diNr\n13ryxhtWhgzJd3WVCvltFjcRESl7utJ2MyYTvPpqDjVqFDBlijfffKN/QhGRqkLf+G6oTh0bL7+c\ng8UCQ4f6cv68q2skIiLOoKDtpsLDrYwencfRox7ExFS8/m0RESl7Ctpu7Kmn8mjd2sK6dZ68+mrF\nvX9bRETKhoK2GzOZIDExh5o1C5g2zZsdO/TPKSJSmelb3s3VqmVj8eIcrFZ45BFfzp51dY1ERKS8\nKGhXAu3aWRk3Lo+ffvLgiSd8KdAEZCIilZKCdiUxZkwe7dpZ+OgjE4sWqX9bRKQyUtCuJIxGWLQo\nh+DgAmbM8GbbNqOrqyQiImVMQbsSCQ628eqrOdhsMGyYD6dPG1xdJRERKUMK2pVM69ZWxo/P48QJ\nD0aM8FH/tohIJaKgXQnFxOTRsaOFzZtNvPiil6urIyIiZURBuxLy8ICXX86hTp0C4uO9SE1V/7aI\nSGWgoF1J1axpIzExB4PB3r+dlqb+bRERd1eqR3POnDmT3bt3YzAYiI2NpUWLFo7XTpw4wZgxY8jP\nzycsLIxnn32WlStX8v777zvW+e6779i1axcDBgwgKysLs9kMwPjx42nevHkZn5L85o47rEycmMdz\nz3nz+OM+vPtuNkZddIuIuK0Sg/b27ds5cuQISUlJHDp0iNjYWJKSkhyvz5o1i8GDBxMZGcm0adM4\nfvw4vXv3pnfv3o7t169f71g/Pj6e0NDQcjgVKcrIkXls22Zk40YT8+d7MXZsnqurJCIiV6nE9Hhq\naioREREANGrUiPPnz5ORkQFAQUEBO3bsIDw8HIC4uDhCQkIKbf/yyy/z+OOPl3W9pZQ8PODFF7Op\nW7eA55/34rPPdKktIuKuSgza6enpBAQEOJYDAwNJS0sD4MyZM/j5+REfH0+/fv1ISEgotO3//vc/\n6tSpQ1BQkKNs4cKFPPDAA0yZMoWcnJyyOg+5jMBAeO01e2r80Ud9OHlS/dsiIu6oVH3af2T7w4Ob\nbTYbJ0+eZODAgdStW5dhw4axZcsWOnToAMCqVauIjo52rD9w4ECaNGlC/fr1iYuLY8WKFQwZMqTY\nYwUEmDGZdGUIEBTkf03bd+0Ks2fD2LEGRo2qxsaNqH+7CNfazlI6amfnUDs7j7PausSgHRwcTHp6\numP51KlTjivngIAAQkJCqF+/PgCtWrXi4MGDjqC9bds2Jk2a5Ng2MjLS8Xd4eDjr1q277LHPns0q\n/ZlUYkFB/qSlXbzm/Tz4IGzc6MOGDZ48/XQuEyaof/uPyqqd5fLUzs6hdnaesm7ry/0AKDE93qZN\nG1JSUgDYs2cPwcHBVKtWDQCTyUS9evU4fPiw4/UGDRoAcPLkSfz8/PDysk/uYbPZGDRoEBcuXADs\nAb1x48ZXf1ZyxQwGWLgwh/r1C3jhBS8++USX2gDJySbatzdjMkH79maSk684ASUi4hQlfju1bNmS\nZs2a0bdvXwwGA3FxcaxevRp/f38iIyOJjY1lwoQJ2Gw2QkNDHYPS0tLSCAwMdOzHYDDQp08fBg0a\nhK+vL7Vq1eKJJ54ovzOTIl1/vb1/+667zIwY4cPmzVnUrm0recNKKjnZxPDhvo7lffuMvy5nEx1t\ncV3FRESKYLD9sZO6glFqx6480lyvvebJM8/4cMcdFlavzsZURS8u27c3s2/fpRmHsDArW7aoe6Y8\nKG3rHGpn56lQ6XGpnIYOzeeuu/L5739NzJ5ddecnP3Cg6I9AceUiIq6kb6YqymCA+fNzuOGGAhYs\n8Objj6tm/3ZoaNGPQSuuXETElRS0q7Dq1WHJkmy8vGyMGOHDzz9Xvfu3R48uegR9TIxG1otIxaOg\nXcW1aFHA9Om5nDnjwSOP+JKf7+oaOVd0tIXExGzCwqyYTPa+7MREDUITkYpJQVsYNCifXr3y+fpr\nIzNmeLu6Ok4XHW1hy5Ys8vNhy5YsBWwRqbAUtAWDARIScmjYsIBFi7xISama/dsiIhWdgrYA4O8P\nr7+ejY+PjSee8OXYsarXvy0iUtEpaItD8+YFzJiRy7lzBh55xJc8jcUSEalQFLSlkAcfzOfee/PZ\nudPI9OlVr39bRKQiU9CWQgwGmDMnh8aNrSQmevHhh1V0qjQRkQpIQVsuUa0avP56Dr6+NmJifDh8\nWP3bIiIVgYK2FOnGGwuYPTuHCxfs/du5ua6ukYiIKGhLsfr2tdC3bz67dxuZOlX92yIirqagLZc1\na1YOTZtaWbLEizVr1L8tIuJKCtpyWWazvX/bbLbx5JM+/Pij+rdFRFxFQVtKFBpawJw5OWRkGBg6\n1JecHFfXSESkalLQllLp3dvCgw/m8d13RiZNUv+2iIgrKGhLqc2YkUtYmJVly7xYvVr92yIizqag\nLaXm62t//rafn42xY3344Qf1b4uIOJOCtlyRRo1szJuXQ2amgSFDfMnKcnWNRESqDgVtuWLR0RYG\nDcpj3z4jzzyj/m0REWdR0Jar8uyzudx0k5UVK7xISlL/toiIMyhoy1Xx8bE/f9vf38b48T58/73e\nSiIi5U3ftHLVGjSwMX9+DllZBoYO9SEz09U1EhGp3BS05Zr06GFh6NA8vv/eyPjxPthsrq6RiEjl\npaAt1ywuLpe//93Kv//tyTvvqH9bRKS8KGjLNfP2hldfzea662xMmODD3r16W4mIlAd9u0qZuOEG\nGwsW5JCTY+/fzshwdY1ERCopHhsdAAAgAElEQVQf5TKlzHTrZmH48DwSE734xz/88PUtn+MYymki\ntqZN4YEHTHTpYsFoLJ9jiIhci1IF7ZkzZ7J7924MBgOxsbG0aNHC8dqJEycYM2YM+fn5hIWF8eyz\nz7Jt2zZiYmJo3LgxAKGhoUyePJkTJ07w9NNPY7VaCQoKYs6cOXh5eZXPmYlLTJ6cS3q6ga++Kp+o\nV14D3axW2LgRNm70pV69Ah5+OI8HHsgnIKB8jnetkpNNzJ/vxYEDHoSGFjB6dB7R0RZXV0tEylmJ\nQXv79u0cOXKEpKQkDh06RGxsLElJSY7XZ82axeDBg4mMjGTatGkcP34cgH/+858sXLiw0L4WLlxI\n//796dq1K/PmzWPVqlX079+/jE9JXMnLC155xT2f3XnypD9z5uSxcqUnzz7rw5w53tx3Xz5Dh+Zz\n440Frq6eQ3KyieHDf09j7Ntn/HU5W4FbpJIrsU87NTWViIgIABo1asT58+fJ+LXDsqCggB07dhAe\nHg5AXFwcISEhxe5r27ZtdOrUCYCOHTuSmpp6zScgUlaaN4e5c3PZvTuDqVNzCAqy8dZbXrRv78e9\n9/qybp0Jq9XVtYT584vOTi1YoKyVSGVXYtBOT08n4A85wsDAQNLS0gA4c+YMfn5+xMfH069fPxIS\nEhzr/fDDDzz66KP069ePL774AoDs7GxHOrxGjRqO/YhUJNdfD48/ns+2bZm8+WY2bdta+OwzE4MG\n+XL77X68/LIn5865rn4HDhT9sS2uXEQqjyseiGb7Q6eizWbj5MmTDBw4kLp16zJs2DC2bNnCjTfe\nyMiRI+natSvHjh1j4MCBfPTRR8XupzgBAWZMJo0IAggK8nd1FaqEP7fzwIH2/777Dl56CZYt82Da\nNB/mzPFhwAB44glo1sy5dQwLg2+/Larc4DbvE3epp7tTOzuPs9q6xKAdHBxMenq6Y/nUqVMEBQUB\nEBAQQEhICPXr1wegVatWHDx4kA4dOtCtWzcA6tevT82aNTl58iRms5mcnBx8fHw4efIkwcHBlz32\n2bN67iPY3wxpaRddXY1K73LtXKsWTJ8OY8bA22978sYbXiQmepCYCG3bWhg6NJ/OnZ0z6nzkyMJ9\n2r8ZMSKbtLSK36et97NzqJ2dp6zb+nI/AErMp7Vp04aUlBQA9uzZQ3BwMNWqVQPAZDJRr149Dh8+\n7Hi9QYMGvP/++yxZsgSAtLQ0Tp8+Ta1atWjdurVjXx999BFt27a9phMTcbaAABgxIp/t2wunzh96\nyHmp8+hoC4mJ2YSFWTGZbISFWUlM1CA0karAYCtFnnru3Ll8/fXXGAwG4uLi2Lt3L/7+/kRGRnLk\nyBEmTJiAzWYjNDSUqVOnkpWVxbhx47hw4QL5+fmMHDmS9u3bc+rUKcaPH09ubi4hISHEx8fj6elZ\n7HH1K9FOv5id42rbee9eD5Ys8WTVKk+ysw2YzTbHqPOmTSvOqPOKQu9n51A7O48zr7RLFbRdRW84\nO334nONa2/nsWVixwpOlS704dsyexHJ26twd6P3sHGpn56lQ6XERKZ2AABg50p46/9e/srnzzsKp\n80WLXDvqXETcn4K2SBkzGu1Tuq5enc2WLZkMGJBHWpqBqVN9uOWWaowb583+/froiciV0zeHSDkK\nCysgISGXb77JYMqUHAIDbSxb5kW7dvYJWzZsMFaICVtExD0oaIs4wR9T50uX/p46HzjQrNS5iJSa\ngraIE5lM0L178anzp57y5vvv9bEUkaLp20HERX5Lne/alcHkybkEBtp4800v2rZV6lxEiqagLeJi\ngYHwxBN5jtR5mzaFU+evvOLJ+fOurqWIVAQK2iIVxG+p8+TkbD75xJ46P3XKQFycDzffrNS5iCho\ni1RIzZr9Pur8z6nz++7zJSVFqXORqkhBW6QC+2Pq/I037KnzTz81MWCAPXX+zjtX/KA+EXFjCtoi\nbsBkgrvu+j11/uCD9tR5TIwvW7dqflSRqkJBW8TNNGtWwLx5uaxdm4XRaGPMGB8yMlxdKxFxBgVt\nETd1880FjBqVx7FjHkyf7u3q6oiIEyhoi7ixMWPyaNrUytKlXnz+udLkIpWdgraIG/P2hgULcvDw\nsDF6tNLkIpWdgraIm/v73wsYOTKPo0c9mDFDaXKRykxBW6QSGDcuj9BQK0uWePHll0qTi1RWCtoi\nlYCPT+E0eWamq2skIuVBQVukkrj11gIeeyyfw4c9iI9XmlykMlLQFqlEnn46l7/9zcprr3ny3/8q\nTS5S2Shoi1Qivr72NDnA6NE+ZGW5uEIiUqYUtEUqmX/8o4Dhw/P58UcPZs1SmlykMlHQFqmEJkzI\npWHDAhITPdm+XR9zkcpCn2aRSshshvnz7WnymBhfsrNdXCERKRMK2iKV1B13WHnkkXwOHfLg+eeV\nJhepDBS0RSqxiRNz+etfC3jlFU++/lofdxF3p0+xSCXm52cfTV5QYCAmxoecHNfVJTnZRPv2Zkwm\naN/eTHKyyXWVEXFTCtoilVyrVlaGDs3j4EEjc+d6uaQOyckmhg/3Zd8+I1Yr7NtnZPhwXwVukSuk\noC1SBTzzTC716xfw0kte7Nrl/I/9/PlF/1hYsMA1PyJE3FWpfubOnDmT3bt3YzAYiI2NpUWLFo7X\nTpw4wZgxY8jPzycsLIxnn30WgOeff54dO3ZgsVgYPnw4nTt3ZsKECezZs4frr78egCFDhtChQ4ey\nPysRKcTPzz6a/J57zIwa5cOmTVl4O3Fs2oEDRf9QKK5cRIpWYtDevn07R44cISkpiUOHDhEbG0tS\nUpLj9VmzZjF48GAiIyOZNm0ax48f5+jRoxw8eJCkpCTOnj1LdHQ0nTt3BmDMmDF07Nix/M5IRIp0\n551WHn44j6VLvUhI8CI2Ns9pxw4NLWDfvkunVQ0NLXBaHUQqgxJ/5qamphIREQFAo0aNOH/+PBkZ\nGQAUFBSwY8cOwsPDAYiLiyMkJIR//OMfLFiwAIDq1auTnZ2N1Wotr3MQkVKaPNmeJn/xRS9273be\nVe7o0UX/QIiJcd4PB5HKoMRPbXp6OgEBAY7lwMBA0tLSADhz5gx+fn7Ex8fTr18/EhISADAajZjN\nZgBWrVpFu3btMBrtv7KXL1/OwIEDefLJJzlz5kyZn5CIFK9aNZg3Lwer1cCoUT7kOSlmRkdbSEzM\nJizMiskEYWFWEhOziY62OKcCIpXEFQ/dtNlshf4+efIkAwcOpG7dugwbNowtW7Y4+qk3bdrEqlWr\neOONNwDo2bMn119/PTfeeCOvvvoqL730ElOmTCn2WAEBZkwmPakIICjI39VVqBKqQjvfey8MHw6J\niUYSE/35dRhKuRs2zP6fnRHwdc6Bq7Cq8H6uKJzV1iUG7eDgYNLT0x3Lp06dIigoCICAgABCQkKo\nX78+AK1ateLgwYN06NCBzz77jMWLF/P666/j7+/veP034eHhTJ069bLHPntWjygC+5shLe2iq6tR\n6VWldn76afjwQz9mzjTQoUMWN93kvL7lqtTOrqR2dp6ybuvL/QAoMT3epk0bUlJSANizZw/BwcFU\nq1YNAJPJRL169Th8+LDj9QYNGnDx4kWef/55EhMTHSPFAZ544gmOHTsGwLZt22jcuPFVn5SIXD1/\nf9ekyUXk2pR4pd2yZUuaNWtG3759MRgMxMXFsXr1avz9/YmMjCQ2NpYJEyZgs9kIDQ0lPDyclStX\ncvbsWUaPHu3Yz+zZs3nggQcYPXo0vr6+mM1m4uPjy/XkRKR4HTpYefDBPJYv92LBAi+eekqRW6Si\nM9j+2EldwSi1Y6c0l3NUxXa+cAHatfPj1CkDH32URfPm5Z8mr4rt7ApqZ+epUOlxEam8qle3p8kt\nFvvc5Pn5rq6RiFyOgrZIFRcebqVfv3y+/dbIiy9qWlGRikxBW0R49tkcatcuICHBi7179bUgUlHp\n0ykiXHcdJCTkkJ9vT5NbNOeJSIWkoC0iAERGWrn//nx27zby8stKk4tURAraIuIwfXoOtWoVMGeO\nF/v36+tBpKLRp1JEHK6/HubOzSEvT2lykYpIQVtECunSxcp99+Wza5eRRYuUJhepSBS0ReQSM2bk\nEBRkT5MfOKCvCZGKQp9GEblEQADMmZNLbq49TW61urpGIgIK2iJSjG7dLNxzTz47dhhZvNjT1dUR\nERS0ReQyZszIpWbNAmbP9uaHHwyuro5IlaegLSLFqlHDxvPP55KTYyAmxldpchEXU9AWkcu66y4L\nPXvm89VXRl57TWlyEVdS0BaREsXH51KjRgEzZ3rz449Kk4u4ioK2iJSoZk0bs2f/lib3oaD8H7st\nIkVQ0BaRUrn7bgt33ZXPtm0mlixRmlzEFRS0RaTUZs3KJTCwgOeeU5pcytbJkwZee82T06f1vroc\nBW0RKbXgYBvx8blkZxt48kmlyaVspKYa6dTJzDPP+NCxo5nPPze6ukoVloK2iFyRXr0sdOuWT2qq\niaVLlSaXq2ezwaJFntxzjy+nTxu455580tIM3HuvL7NmeemBNUVQ0BaRK2IwwOzZuQQE2Jg+3ZvD\nh5XOlCt38SIMHuzD1Kk+1KxpIzk5m8WLc1i7Not69WzMm+fN3XebOXpU768/UtAWkStWq5aNGTNy\nyMoyMGaM0uRyZfbt8yAy0o8PP/SkdWsLmzZlcccd9pl7brutgM2bM4mOzufrr42Eh/vx/vsmF9e4\n4lDQFpGrcu+9FqKi8vn8cxPLlilNLqWzcqWJrl3N/PijByNH5rJqVTa1atkKrVO9OixenMOCBdlY\nLDB0qC9jxniTmemiSlcgCtoiclUMBvuTwK67zsa0ad5KY8pl5ebC0097M2KEL0YjLF2azZQpeZiK\nuYg2GKBfPwubNmXSvLmV5cu96NzZzHffVe2wVbXPXkSuSa1aNp57LofMTPtocput5G3cSXKyifbt\nzdSpU4327c0kJytNezV++snA3Xeb+de/vAgLs7JxYybdu5dulNnf/mZj/foshg3L4+BBI127mlmy\nxLPSvddKS0FbRK5Jnz4WIiMtfPaZibfeqjxp8uRkE8OH+7JvnxGr1cC+fUaGD/dV4L5CmzcbiYgw\ns2uXkT598lm3LouGDa8s4np7w3PP5bJ8eRbVqtmYONGHgQN9q+Q93QraInJNDAaYOzeH6tVtTJ3q\nzU8/VY4v0vnzvYosX7Cg6HIprKAA5s71ol8/XzIyDMydm8OLL+ZgNl/9Pjt3tvLJJ1m0bWshJcVU\nJe/pVtAWkWtWp46N6dNzyMiwjyavDKnLAweK/nosrlx+d+YMPPCAL88/781f/mLjgw+yGDgwH0MZ\n/J6rXdvGv/+dzaRJuVXynm69+0SkTPTtayE83MKWLSbeftv90+ShoUXfx1Zcudh98439dq6PPzYR\nHm5h48ZMbrmlbNvMaIRRo/Kq5D3dpQraM2fO5P7776dv377873//K/TaiRMn6NevH/fddx9Tpky5\n7DYnTpxgwIAB9O/fn5iYGPLy8srwVETElQwGSEjIwd/fxpQp3hw/7t5foKNHF/39FBOj762i2Gyw\nbJknd91l5qefDDz9dC5vv51NYGD5HfO3e7p79ao693SXGLS3b9/OkSNHSEpKYsaMGcyYMaPQ67Nm\nzWLw4MGsWrUKo9HI8ePHi91m4cKF9O/fn7fffpsbbriBVatWlc9ZiYhL1K1r49lnc7l40cDYse6d\nJo+OtpCYmE1YmBWTyUZYmJXExGyio6tIHvYKZGXBqFE+jBvng58fvPNONuPG5eHhhFxu9eqQmFh1\n7ukusUlTU1OJiIgAoFGjRpw/f56MjAwACgoK2LFjB+Hh4QDExcUREhJS7Dbbtm2jU6dOAHTs2JHU\n1NRyOSkRcZ3+/fPp0MHCxx+bSEpy76ue6GgLW7Zkcfx4Blu2ZClgF+HHHw1062YmKcmTv//dyqZN\nmYSHW51ah6p0T3eJZ5Senk5AQIBjOTAwkLS0NADOnDmDn58f8fHx9OvXj4SEhMtuk52djZeXfeRl\njRo1HPsRkcrDYIB583KoVs3GpEk+nDjh3mlyKd769SYiI/3Yu9fIoEF5vP++vY/ZVarCPd1X/DPY\n9oezt9lsnDx5koEDB1K3bl2GDRvGli1bLrvN5cr+LCDAjMlUtYbzFycoyN/VVagS1M5lIygIEhJg\n+HCIja3G2rUUGjmsdnaO8mpniwUmTYLZs8HXF5YtgwEDvICKcTtcYiL06AEPP2xg4kQfvvzShzfe\ngJo1y++YznpPlxi0g4ODSU9PdyyfOnWKoKAgAAICAggJCaF+/foAtGrVioMHDxa7jdlsJicnBx8f\nH06ePElwcPBlj332bNZVnVRlExTkT1raRVdXo9JTO5etXr1gxQpfPvzQxKJF2fTpY08tq52do7za\n+dQpA8OH+/DFFyYaNChg6dJswsIKqGiJ09tvh82bDYwY4cPatSZuuqmARYtyuPPOsk/dl3VbX+4H\nQInp8TZt2pCSkgLAnj17CA4Oplq1agCYTCbq1avH4cOHHa83aNCg2G1at27tKP/oo49o27btNZ2Y\niFRcv6XJzWYbzzzjw8mTSpO7u//+10inTma++MJEt275bNyYSVhYxb0FrjLe022wlSJPPXfuXL7+\n+msMBgNxcXHs3bsXf39/IiMjOXLkCBMmTMBmsxEaGsrUqVPx8PC4ZJumTZty6tQpxo8fT25uLiEh\nIcTHx+PpWfz9nPo1bqcrE+dQO5ePpUs9GT/eh6iofN58M4fgYLWzM5Tl+9lmg8RET6ZN8wZg0qRc\nHn+8bCZLcZavv/bg0Ud9OXrUg9tus7J4cTb165dNZ7czr7RLFbRdRR9sOwUT51A7l4+CArjvPl8+\n/9zEK69k8+ijvmpnJyir93NGBsTE+LB2rSdBQQW89loOrVs7d3R4WblwAcaN8+G99zypXt3GvHk5\n3H33tV92V6j0uIjItfDwgBdesKfJY2N9+OUXV9dISmv/fg86dzazdq0nt99uYfPmLLcN2FA57ulW\n0BaRcnfDDTYmT87l7FkDw4bZn60sFdvq1Saiosz88IORxx7LY/XqbGrVqrCJ2VIr7p7uPXvcIxy6\nRy1FxO09/HA+bdpYWLsWwsPN/Pe/up2zIsrLg4kTvXn0UV88PGDJkmymTcvlMsOP3NKf7+mOinKP\ne7oVtEXEKTw8YNmybEaMgB9+8ODuu82MHevNuXOurpn85uefDfTsaWbJEi9uvNHKxo2Z9OjhxkOt\nS1DUc7ofesiHM2dcXbPiKWiLiNP4+8NLL8GHH2Zx441W3nrLizZt/FizxlThr3Aqu61bjUREmNmx\nw8i99+azbl0WjRpVjX+UPz6ne8MGTzp29OOLLypmJkhBW0Sc7rbbCti0KYtnnsnlwgUDjzziy4MP\n+nLsmBvdQ1RJFBTACy940aePLxcuGJg9O4dFi3Lw83N1zZzrj/d0nzpl4J57KuY93QraIuISnp72\nx1xu3ZpJ27YWNm400batH4sXe1a4L8rK6tw5GDDAl/h4b0JCbLz/fhYPP+xe91+XpaKe092zp7lC\n/ZhU0BYRl2rY0MaqVdm8+GI2Pj42pkzxoWtXM99+q6+n8vS//3kQEeHHxo0m2re3sGlTFrfeWnFn\nN3OmPz6n+6uvjHTsWHGe061PhYi4nMEA999v4YsvsujTJ5/du41ERpqJi3Ove2jdxYoVnnTvbubo\nUQ/Gjs3l3XezqVGjavRfl9Zv93TPn//7Pd1jx3qT5eJHYihoi0iFUaOGjZdeymHlyizq17fxyite\ntGvnx8cfV8xBQe4mOxtGj/bmySd98PWFt9/OYvz4PIxq3iIZDNC/v4WNG7No1sw+cNLV93QraItI\nhdO+vZWtWzOJicnlxAkD/fqZGT7ch1OnKk7fors5fNhA9+5m3n7bi5tvtrJpUyYREe47u5kzNW5c\n4Lin+8AB197TraAtIhWSry8880zer32tVpKTPWnTxo/lyz0pUNfrFUlJMRIR4cd33xkZMMA+0Kqs\nHpZRVfj4/H5Pt5+f6+7pVtAWkQotLKyADz7IIj4+B6sVxozxITral4MH9fVVEosFZszwYsAAM3l5\nsHBhNgkJufj4uLpm7qtzZytbtmRx552/39N96JDzjq93vYhUeEYjDBmSz+efZ9KtWz6pqSY6djQz\nZ46X5jEvxqlTcP/9vixY4M1f/1rAhx9m0bev7qUrC7Vr21i5MptnnsklM9PAkSPOO7aCtoi4jZAQ\nG//6Vw5Ll2YTGGhjzhxvzWP+J2fOwFtvedKyJXz2mYmoqHw2bszkppvUp1CWjEb7PAMHD2YQHu68\n4ypoi4jb6d7dwhdfZDJkSJ7mMcf+zOuVK0088IAvzZtXY+xYH06cgEmTcvnXv3K47jpX17DycvZE\nNAraIuKW/P0hPj73knnM33uvasxjnp0NH3xgYsgQH8LCqjFihC8bN5oICbERHGy/qv7Pf0ysWVMx\nJgWRsmGw2Sru2zst7aKrq1AhBAX5qy2cQO3sHOXRzvn58MorXsyd60VOjoGICAuzZ+dQr16F/Xq7\nKvn58OmnRpKTPVm3zkRGhv0yr3FjK9HRFvz8bMTFXTrKLDExm+ho9WeXl7J+TwcF+Rf7moK2G1Aw\ncQ61s3OUZzv/3/8ZeOopHz791ITZbGP8+FweeSQfkxtfbBYUwH//a2T1ahMffGDizBl7grRevQJ6\n9conOtpCs2YFGAzQvr2Zffsu7d8PC7OPeJbyoaD9K32B2imYOIfa2TnKu51tNnv/7pQp3pw540GL\nFlbmzcuhRQv3GYhls8E333iwerUna9aY+OUXe6AOCiqgVy8LvXrlc9ttBZf0p9apUw2r9dJOVpPJ\nxvHjGc6oepXkzKDtxr8/RUQuZTBAnz4WOnWyMnWqN0lJnnTubGbYsHyefjqXatVcXcPi7dvnwXvv\nmUhO9uTwYXugvv56Gw8+mEevXhbatLFedsrR0NCCIq+0Q0Pd5weLXJ6CtohUSjVq2HjxxRx6985n\n3DgfFi/24sMPTcyenVOhpu/8v/8z8N57nrz3nskRcM1mG/fck8899+TToYMVL6/S7Wv06DyGD/e9\npDwmJq8sqywupKAtIpVau3b2ecxfeMGLl17yon9/M7165TN9ei61armmd/DECQNr1tivqHftsgdq\nb28b3brZ+6gjIy2YzVe+X/tgs2wWLPDiwAEjoaFWYmLyNAitElGfthtQX6tzqJ2dw5XtvHevB2PH\n+rBjh5HrrrMRF5dL//75eDjh5tfTpw2sXWvivfdMpKYasdkMGI022rWzEh2dT7duFqpXL7vj6f3s\nPOrTFhEpB7/NY/7mm54895w3Y8b48O9/m0hIyKVx47Lv9714EdatM/Hee55s3WrEYrEPErvjDgvR\n0RZ69LBQs2aFvW6SCkhBW0SqFKMRBg/Op2tXC7Gx3nz4oScdOxoZNSqPmJg8vL2vbf/Z2bBpk4nV\nq01s2mQiN9ceqG+5xUqvXvn06mUhJESBWq6OgraIVEl16thYujSHdessTJzozdy53qxZY2Lu3Fxa\ntbqygWp5ebB1q33Sk/XrTWRm2gN1kyb2SU969cqnYUMFarl2CtoiUqV162ahbVsL8fHeLFniSc+e\nZh58MI8pU3K5/vrit7NaITXVSHKyiQ8+8OTsWXugrl+/gKFD7YO/brzx0nupRa6FgraIVHn+/jBz\nZi733pvPmDE+LF/uRUqKiRkzcunZ0+IIvDYb7NzpQXKyfdKTkyftI9hq1Spg+PB8evXKp2VLBWop\nP6UK2jNnzmT37t0YDAZiY2Np0aKF47Xw8HBq166N8dc7/ufOncunn37K+++/71jnu+++Y9euXQwY\nMICsrCzMv97LMH78eJo3b16W5yMictVuvbWATZuyHPOYDxvmS1KShccfz3PM+X30qD1QBwTYGDAg\nj3vusXDHHZef9ESkrJQYtLdv386RI0dISkri0KFDxMbGkpSUVGid1157DT8/P8dy79696d27t2P7\n9evXO16Lj48nNDS0rOovIlKmPD1h1Kg8evTI56mnfPj4YxMff2z/qvTzs9G7dz7R0fm0a1f6SU9E\nykqJQTs1NZWIiAgAGjVqxPnz58nIyKBaKecCfPnll5k7d+611VJExMkaNLCxcmU2q1eb+PJLIx06\nWImIsOB76YRjIk5TYtBOT0+nWbNmjuXAwEDS0tIKBe24uDh+/vlnbr31VsaOHYvh1w6d//3vf9Sp\nU4egoCDHugsXLuTs2bM0atSI2NhYfHwufYyciEhFYDDAvfdauPdezSgmFcMVD0T78wRqo0aNom3b\ntlx33XWMGDGClJQUoqKiAFi1ahXR0dGOdQcOHEiTJk2oX78+cXFxrFixgiFDhhR7rIAAMyaTOorg\n8jPkSNlROzuH2tk51M7O46y2LjFoBwcHk56e7lg+depUoSvnXr16Of5u164dBw4ccATtbdu2MWnS\nJMfrkZGRjr/Dw8NZt27dZY999qye/wqajtBZ1M7OoXZ2DrWz8zhzGtMSZ9xt06YNKSkpAOzZs4fg\n4GBHavzixYsMGTKEvDz7E2S++uorGjduDMDJkyfx8/PD69eRGjabjUGDBnHhwgXAHtB/W1dERERK\nVuKVdsuWLWnWrBl9+/bFYDAQFxfH6tWr8ff3JzIyknbt2nH//ffj7e1NWFiY4yo7LS2NwMBAx34M\nBgN9+vRh0KBB+Pr6UqtWLZ544onyOzMREZFKRk/5cgNKczmH2tk51M7OoXZ2ngqVHhcREZGKQUFb\nRETETShoi4iIuAkFbRERETehoC0iIuImFLRFRETchIK2iIiIm1DQFhERcRMK2iIiIm5CQVtERMRN\nKGiLiIi4CQVtERERN6GgLSIi4iYUtEVERNyEgraIiIibUNAWERFxEwraIiIibkJBW0REXCo52UT7\n9mbq1KlG+/ZmkpNNrq5ShaWWERERl0lONjF8uK9jed8+46/L2URHW1xXsQpKV9oiIuIy8+d7FVm+\nYEHR5VWdgraIiLjMgdTFykAAAAgNSURBVANFh6Hiyqs6tYqIiLhMaGjBFZVXdQraIiLiMqNH5xVZ\nHhNTdHlVp6AtIiIuEx1tITExm7AwKyaTjbAwK4mJGoRWHI0eFxERl4qOtihIl5KutEVERNyEgraI\niIibUNAWERFxEwraIiIibkJBW0RExE2UavT4zJkz2b17NwaDgdjYWFq0aOF4LTw8nNq1a2M0GgGY\nO3cuhw8fJiYmhsaNGwMQGhrK5MmTOXHiBE8//TRWq5WgoCDmzJmDl5emqhMRESmNEoP29u3bOXLk\nCElJSRw6dIjY2FiSkpIKrfPaa6/h5+fnWD58+DD//Oc/WbhwYaH1Fi5cSP/+/enatSvz5s1j1apV\n9O/fv4xORUREpHIrMT2emppKREQEAI0aNeL8+fNkZGRc1cG2bdtGp06dAOjYsSOpqalXtR8REZGq\nqMQr7fT0dJo1a+ZYDgwMJC0tjWrVqjnK4uLi+Pnnn7n11lsZO3YsAD/88AOPPvoo58+fZ+TIkbRp\n04bs7GxHOrxGjRqkpaVd9tgBAWZMJuNVnVhlExTk7+oqVAlqZ+dQOzuH2tl5nNXWVzwjms1mK7Q8\natQo2rZty3XXXceIESNISUnh73//OyNHjqRr164cO3aMgQMH8tFHH112P0U5ezbrSqtXKQUF+ZOW\ndtHV1aj01M7OoXZ2DrWz85R1W1/uB0CJ6fHg4GDS09Mdy6dOnSIoKMix3KtXL2rUqIHJZKJdu3Yc\nOHCAWrVq0a1bNwwGA/Xr1///9u4nJKotgOP4z+cw+casZGKMgZIWRdimghZOakVKQRBa1MigbVoU\nUegiUCRIkIKRFhpFStqmECYmp1wEMyQNuNAIgqAorFalNUmNJOpg/nmLQgh89Ox5vXNv389uzmLO\n796FP885l7lau3atEomEXC6XUqmUJCmRSMjj8fyf6wIA4I/yy9LetWuXotGoJOnFixfyeDzzW+Nj\nY2M6ceKEpqa+v43lyZMn2rRpk3p6etTZ2SlJGhkZ0efPn5WXlyefzzf/XbFYTMXFxYZcFAAAdvTL\n7fEdO3Zo69atqqysVEZGhi5cuKDu7m7l5OSorKxMJSUl8vv9WrFihQoKCnTgwAGNj4/r3Llz6u3t\n1bdv39TY2Cin06mzZ8+qrq5OoVBIXq9X5eXly3GNAADYQsbcfzlcNgnnMd9xNrU8uM/Lg/u8PLjP\nyyetzrQBAEB6oLQBALAIShsAAIugtAEAsAhKGwAAi6C0AQCwCEobAACLoLQBALAIShsAgN8QiTi0\ne7dLDoe0e7dLkcii38G1aMbPAACAzUQiDp08+ff855cvM398nlRFxbRh87LSBgBgkVpanAuOt7Yu\nPL5UKG0AABZpcHDh+vy38aVCaQMAsEibN88uanypUNoAACxSbe3UguM1NQuPLxVKGwCARaqomFZ7\n+6QKCmbkcEgFBTNqbzf2ITSJp8cBAPgtFRXTqqiY/vE+7YllmZOVNgAAFkFpAwBgEZQ2AAAWQWkD\nAGARlDYAABZBaQMAYBGUNgAAFkFpAwBgEZQ2AAAWkTE3NzdndggAAPBrrLQBALAIShsAAIugtAEA\nsAhKGwAAi6C0AQCwCEobAACLoLTTXHNzs/x+v44cOaJYLGZ2HNtKpVIqLS1Vd3e32VFsraenR4cO\nHdLhw4cVj8fNjmNL4+PjOnPmjKqrq1VZWam+vj6zI9nO4OCgSktLdfv2bUnShw8fVF1drUAgoJqa\nGk1NTRk2N6WdxgYGBvT69WuFQiF1dHTo0qVLZkeyrevXr2v16tVmx7C1ZDKpa9euqaurS21tbert\n7TU7ki1FIhFt3LhRt27dUmtrqy5evGh2JFuZmJhQU1OTCgsL58euXLmiQCCgrq4u5efnKxwOGzY/\npZ3Gdu7cqdbWVknSqlWrNDk5qZmZGZNT2c/bt2/15s0b7dmzx+wottbf36/CwkKtXLlSHo9HTU1N\nZkeypdzcXI2OjkqSvn79qtzcXJMT2YvT6dSNGzfk8Xjmxx4/fqx9+/ZJkvbu3av+/n7D5qe001hm\nZqZcLpckKRwOq6SkRJmZmSansp9gMKj6+nqzY9je+/fvlUqldOrUKQUCAUP/sP3JDh48qOHhYZWV\nlamqqkp1dXVmR7IVh8OhrKysn8YmJyfldDolSW63WyMjI8bNb9g3Y8k8fPhQ4XBYN2/eNDuK7dy7\nd0/btm3T+vXrzY7yRxgdHdXVq1c1PDys48eP69GjR8rIyDA7lq3cv39fXq9XnZ2devXqlRoaGnhW\nYxkZ/cvglHaa6+vrU1tbmzo6OpSTk2N2HNuJx+N69+6d4vG4Pn78KKfTqXXr1snn85kdzXbcbre2\nb98uh8OhDRs2KDs7W1++fJHb7TY7mq08ffpURUVFkqQtW7bo06dPmpmZYZfOQC6XS6lUSllZWUok\nEj9tnS81tsfT2NjYmJqbm9Xe3q41a9aYHceWWlpadPfuXd25c0dHjx7V6dOnKWyDFBUVaWBgQLOz\ns0omk5qYmOC81QD5+fl69uyZJGloaEjZ2dkUtsF8Pp+i0agkKRaLqbi42LC5WGmnsQcPHiiZTKq2\ntnZ+LBgMyuv1mpgK+D15eXnav3+/jh07Jkk6f/68/vqLdcNS8/v9amhoUFVVlaanp9XY2Gh2JFt5\n/vy5gsGghoaG5HA4FI1GdfnyZdXX1ysUCsnr9aq8vNyw+Xk1JwAAFsG/uQAAWASlDQCARVDaAABY\nBKUNAIBFUNoAAFgEpQ0AgEVQ2gAAWASlDQCARfwDau4PTEkMtcIAAAAASUVORK5CYII=\n","text/plain":["<matplotlib.figure.Figure at 0x7f53e46e39b0>"]},"metadata":{"tags":[]}}]},{"metadata":{"id":"jo09kaTgYY88","colab_type":"text"},"cell_type":"markdown","source":["\n","The model quickly starts overfitting, unsurprisingly given the small number of training samples. Validation accuracy has high variance for \n","the same reason, but seems to reach high 50s.\n","\n","Note that your mileage may vary: since we have so few training samples, performance is heavily dependent on which exact 200 samples we \n","picked, and we picked them at random. If it worked really poorly for you, try picking a different random set of 200 samples, just for the \n","sake of the exercise (in real life you don't get to pick your training data).\n","\n","We can also try to train the same model without loading the pre-trained word embeddings and without freezing the embedding layer. In that \n","case, we would be learning a task-specific embedding of our input tokens, which is generally more powerful than pre-trained word embeddings \n","when lots of data is available. However, in our case, we have only 200 training samples. Let's try it:"]},{"metadata":{"id":"APu9pzKHYY89","colab_type":"code","outputId":"29da735e-5398-454a-a14d-9059b0179bf3","executionInfo":{"status":"ok","timestamp":1540394379030,"user_tz":-120,"elapsed":124608,"user":{"displayName":"Krzysztof Suwada","photoUrl":"","userId":"18334398126529369335"}},"colab":{"base_uri":"https://localhost:8080/","height":622}},"cell_type":"code","source":["from keras.models import Sequential\n","from keras.layers import Embedding, Flatten, Dense\n","\n","model = Sequential()\n","model.add(Embedding(max_words, embedding_dim, input_length=maxlen))\n","model.add(Flatten())\n","model.add(Dense(32, activation='relu'))\n","model.add(Dense(1, activation='sigmoid'))\n","model.summary()\n","\n","model.compile(optimizer='rmsprop',\n","              loss='binary_crossentropy',\n","              metrics=['acc'])\n","history = model.fit(x_train, y_train,\n","                    epochs=10,\n","                    batch_size=32,\n","                    validation_data=(x_val, y_val))"],"execution_count":26,"outputs":[{"output_type":"stream","text":["_________________________________________________________________\n","Layer (type)                 Output Shape              Param #   \n","=================================================================\n","embedding_2 (Embedding)      (None, 100, 100)          1000000   \n","_________________________________________________________________\n","flatten_2 (Flatten)          (None, 10000)             0         \n","_________________________________________________________________\n","dense_4 (Dense)              (None, 32)                320032    \n","_________________________________________________________________\n","dense_5 (Dense)              (None, 1)                 33        \n","=================================================================\n","Total params: 1,320,065\n","Trainable params: 1,320,065\n","Non-trainable params: 0\n","_________________________________________________________________\n","Train on 20000 samples, validate on 5000 samples\n","Epoch 1/10\n","20000/20000 [==============================] - 13s 632us/step - loss: 0.4307 - acc: 0.7904 - val_loss: 0.3535 - val_acc: 0.8486\n","Epoch 2/10\n","20000/20000 [==============================] - 12s 618us/step - loss: 0.1382 - acc: 0.9505 - val_loss: 0.4615 - val_acc: 0.8238\n","Epoch 3/10\n","20000/20000 [==============================] - 12s 612us/step - loss: 0.0161 - acc: 0.9957 - val_loss: 0.6569 - val_acc: 0.8258\n","Epoch 4/10\n","20000/20000 [==============================] - 12s 607us/step - loss: 9.0622e-04 - acc: 0.9996 - val_loss: 0.8654 - val_acc: 0.8270\n","Epoch 5/10\n","20000/20000 [==============================] - 12s 607us/step - loss: 8.4787e-06 - acc: 1.0000 - val_loss: 1.1034 - val_acc: 0.8240\n","Epoch 6/10\n","20000/20000 [==============================] - 12s 598us/step - loss: 2.6501e-07 - acc: 1.0000 - val_loss: 1.1343 - val_acc: 0.8266\n","Epoch 7/10\n","20000/20000 [==============================] - 13s 649us/step - loss: 1.2536e-07 - acc: 1.0000 - val_loss: 1.1545 - val_acc: 0.8272\n","Epoch 8/10\n","20000/20000 [==============================] - 13s 638us/step - loss: 1.1030e-07 - acc: 1.0000 - val_loss: 1.1602 - val_acc: 0.8270\n","Epoch 9/10\n","20000/20000 [==============================] - 12s 611us/step - loss: 1.0975e-07 - acc: 1.0000 - val_loss: 1.1616 - val_acc: 0.8274\n","Epoch 10/10\n","20000/20000 [==============================] - 12s 615us/step - loss: 1.1001e-07 - acc: 1.0000 - val_loss: 1.1650 - val_acc: 0.8260\n"],"name":"stdout"}]},{"metadata":{"id":"a-X3OAVWYY9A","colab_type":"code","colab":{}},"cell_type":"code","source":["acc = history.history['acc']\n","val_acc = history.history['val_acc']\n","loss = history.history['loss']\n","val_loss = history.history['val_loss']\n","\n","epochs = range(1, len(acc) + 1)\n","\n","plt.plot(epochs, acc, 'bo', label='Training acc')\n","plt.plot(epochs, val_acc, 'b', label='Validation acc')\n","plt.title('Training and validation accuracy')\n","plt.legend()\n","\n","plt.figure()\n","\n","plt.plot(epochs, loss, 'bo', label='Training loss')\n","plt.plot(epochs, val_loss, 'b', label='Validation loss')\n","plt.title('Training and validation loss')\n","plt.legend()\n","\n","plt.show()"],"execution_count":0,"outputs":[]},{"metadata":{"id":"Rv9q0TpkYY9D","colab_type":"text"},"cell_type":"markdown","source":["\n","Validation accuracy stalls in the low 50s. So in our case, pre-trained word embeddings does outperform jointly learned embeddings. If you \n","increase the number of training samples, this will quickly stop being the case -- try it as an exercise.\n","\n","Finally, let's evaluate the model on the test data. First, we will need to tokenize the test data:"]},{"metadata":{"id":"OOH5Sl8_YY9D","colab_type":"code","colab":{}},"cell_type":"code","source":["test_dir = os.path.join(imdb_dir, 'test')\n","\n","labels = []\n","texts = []\n","\n","for label_type in ['neg', 'pos']:\n","    dir_name = os.path.join(test_dir, label_type)\n","    for fname in sorted(os.listdir(dir_name)):\n","        if fname[-4:] == '.txt':\n","            f = open(os.path.join(dir_name, fname))\n","            texts.append(f.read())\n","            f.close()\n","            if label_type == 'neg':\n","                labels.append(0)\n","            else:\n","                labels.append(1)\n","\n","sequences = tokenizer.texts_to_sequences(texts)\n","x_test = pad_sequences(sequences, maxlen=maxlen)\n","y_test = np.asarray(labels)"],"execution_count":0,"outputs":[]}]}