第16章 循环神经网络算法.ipynb

{
 "cells": [
  {
   "attachments": {
    "LSTM.png": {
     "image/png": "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"
    },
    "RNN.png": {
     "image/png": "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"
    }
   },
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "循环神经网络（Recurrent Neural Networks，RNN）算法是深度学习算法中非常有名的一种算法。RNN之所以称为循环神经网络，是因为一个序列当前的输出与前面的输出有关。具体的表现形式为：网络会对前面的信息进行记忆并应用于当前输出的计算中，即隐藏层之间的节点不再无连接而是有连接的，并且隐藏层的输入不仅包括输入层的输出还包括上一时刻隐藏层的输出。理论上，RNN能够对任何长度的序列数据进行处理。但是在实践中，为了降低复杂性往往假设当前的状态只与前面的几个状态相关。\n",
    "\n",
    "RNN的独特能力来自于它特殊的结构，x代表输入，h代表输出，输出的一部分会作为输入的一部分重新输入，于是RNN具有了一定的记忆性。\n",
    "\n",
    "![RNN.png](attachment:RNN.png)\n",
    "\n",
    "但是生活的经验告诉我们，在比较复杂的情况下，只分析时序数据的最近几个数据是难以得到合理的结果的，需要更长的记忆来追根溯源，于是就有了LSTM（Long Short Term Memory），LSTM可以在更长的时间范围内分析时序数据。\n",
    "\n",
    "![LSTM.png](attachment:LSTM.png)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-25T11:30:12.788144Z",
     "start_time": "2018-11-25T11:29:37.380900Z"
    }
   },
   "outputs": [],
   "source": [
    "import tflearn\n",
    "from tflearn.datasets import mnist\n",
    "import os\n",
    "os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2' \n",
    "os.environ['KMP_DUPLICATE_LIB_OK'] = 'True'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-24T17:13:49.461263Z",
     "start_time": "2018-11-24T17:13:48.013598Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Extracting mnist/train-images-idx3-ubyte.gz\n",
      "Extracting mnist/train-labels-idx1-ubyte.gz\n",
      "Extracting mnist/t10k-images-idx3-ubyte.gz\n",
      "Extracting mnist/t10k-labels-idx1-ubyte.gz\n"
     ]
    }
   ],
   "source": [
    "X, Y, testX, testY = mnist.load_data(one_hot=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-24T17:13:49.472602Z",
     "start_time": "2018-11-24T17:13:49.465078Z"
    }
   },
   "outputs": [],
   "source": [
    "X = X.reshape(-1, 28, 28)\n",
    "testX = testX.reshape(-1, 28, 28)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-24T17:13:51.319337Z",
     "start_time": "2018-11-24T17:13:49.477778Z"
    }
   },
   "outputs": [],
   "source": [
    "net = tflearn.input_data(shape=[None, 28, 28])\n",
    "net = tflearn.lstm(net, 128, return_seq=True)\n",
    "net = tflearn.lstm(net, 128)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-24T17:13:51.362287Z",
     "start_time": "2018-11-24T17:13:51.322125Z"
    }
   },
   "outputs": [],
   "source": [
    "net = tflearn.fully_connected(net, 10, activation='softmax')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-24T17:13:51.544903Z",
     "start_time": "2018-11-24T17:13:51.366680Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From /Users/johnnie/anaconda3/lib/python3.6/site-packages/tflearn/objectives.py:66: calling reduce_sum (from tensorflow.python.ops.math_ops) with keep_dims is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "keep_dims is deprecated, use keepdims instead\n"
     ]
    }
   ],
   "source": [
    "net = tflearn.regression(net, optimizer='adam', loss='categorical_crossentropy', name='output1')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-24T17:14:00.415145Z",
     "start_time": "2018-11-24T17:13:51.557146Z"
    }
   },
   "outputs": [],
   "source": [
    "# 创建神经网络实体\n",
    "model = tflearn.DNN(net, tensorboard_verbose=2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-24T17:21:25.769823Z",
     "start_time": "2018-11-24T17:14:00.419994Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training Step: 773  | total loss: \u001b[1m\u001b[32m0.25896\u001b[0m\u001b[0m | time: 426.972s\n",
      "| Adam | epoch: 001 | loss: 0.25896 - acc: 0.9257 -- iter: 49472/49500\n",
      "Training Step: 774  | total loss: \u001b[1m\u001b[32m0.25194\u001b[0m\u001b[0m | time: 441.524s\n",
      "| Adam | epoch: 001 | loss: 0.25194 - acc: 0.9285 | val_loss: 0.21686 - val_acc: 0.9367 -- iter: 49500/49500\n",
      "--\n"
     ]
    }
   ],
   "source": [
    "model.fit(X, Y, n_epoch=1, validation_set=0.1, show_metric=True, snapshot_step=100)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "其中主要参数的含义为：\n",
    "\n",
    "* n_epoch，整个数据集合训练的次数；\n",
    "* validation_set，验证数据集的比例，也可以直接填写集合，比如（testX，testY），\n",
    "* show_metric，是否展现完整训练过程，\n",
    "* snapshot_step，snapshot的训练步长。"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "识别恶意评论\n",
    "\n",
    "从早期的BBS到现在的贴吧、微信和微博，恶意评论一直都挥之不去，有的是网友的无意吐槽，有的是水军的恶意攻击。以往大型互联网企业都需要雇佣大量的运营支撑人员，人工进行筛选，随着互联网的发展，尤其是移动互联网的发展，人工维护几乎成了不可能完成的任务。这次我们尝试使用机器学习的方法来识别恶意评论"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-02-18T13:49:56.565786Z",
     "start_time": "2019-02-18T13:49:56.552993Z"
    }
   },
   "outputs": [],
   "source": [
    "import glob\n",
    "import keras\n",
    "import numpy as np\n",
    "import tensorflow as tf\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.naive_bayes import GaussianNB\n",
    "from sklearn.metrics import accuracy_score\n",
    "\n",
    "from keras.layers.core import Dense\n",
    "from keras.layers.embeddings import Embedding\n",
    "from keras.layers import LSTM\n",
    "from keras import Sequential\n",
    "from keras.preprocessing.sequence import pad_sequences\n",
    "from keras.utils import to_categorical"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-02-18T13:49:58.261115Z",
     "start_time": "2019-02-18T13:49:58.250136Z"
    }
   },
   "outputs": [],
   "source": [
    "def load_data(path, neg=False, pos=True):\n",
    "    fileList = glob.glob(path + '*.txt')\n",
    "    positive = []\n",
    "    negative = []\n",
    "    count = 0\n",
    "    if pos:\n",
    "        for file in fileList:\n",
    "            posi = []\n",
    "            f1 = open(file)\n",
    "            for line in f1.readlines():\n",
    "                posi.append(line.strip('\\n'))\n",
    "                count += 1\n",
    "            positive.append(posi)\n",
    "        return positive, count\n",
    "    if neg:\n",
    "        for file in fileList:\n",
    "            nega = []\n",
    "            f1 = open(file)\n",
    "            for line in f1.readlines():\n",
    "                posi.append(line.strip('\\n'))\n",
    "                count += 1\n",
    "            negativea.append(nega)\n",
    "        return negative, count"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-02-18T13:50:14.133860Z",
     "start_time": "2019-02-18T13:50:01.242362Z"
    }
   },
   "outputs": [],
   "source": [
    "positive, labelsize1 = load_data('./数据集/movie-review-data/review_polarity/txt_sentoken/pos/')\n",
    "y1 = [0] * labelsize1\n",
    "negative, labelsize2 = load_data('./数据集/movie-review-data/review_polarity/txt_sentoken/neg/')\n",
    "y2 = [1] * labelsize2\n",
    "sentiment = positive + negative\n",
    "y = y1 + y2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-02-18T13:50:18.463898Z",
     "start_time": "2019-02-18T13:50:16.160668Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From <ipython-input-13-ecce7937d04f>:5: VocabularyProcessor.__init__ (from tensorflow.contrib.learn.python.learn.preprocessing.text) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Please use tensorflow/transform or tf.data.\n",
      "WARNING:tensorflow:From /Users/johnnie/anaconda3/lib/python3.7/site-packages/tensorflow/contrib/learn/python/learn/preprocessing/text.py:154: CategoricalVocabulary.__init__ (from tensorflow.contrib.learn.python.learn.preprocessing.categorical_vocabulary) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Please use tensorflow/transform or tf.data.\n",
      "WARNING:tensorflow:From /Users/johnnie/anaconda3/lib/python3.7/site-packages/tensorflow/contrib/learn/python/learn/preprocessing/text.py:170: tokenizer (from tensorflow.contrib.learn.python.learn.preprocessing.text) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Please use tensorflow/transform or tf.data.\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<tensorflow.contrib.learn.python.learn.preprocessing.text.VocabularyProcessor at 0x1a31e77c50>"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sentiments = []\n",
    "for sentences in sentiment:\n",
    "    for sentence in sentences:\n",
    "        sentiments.append(sentence)\n",
    "vp = tf.contrib.learn.preprocessing.VocabularyProcessor(3000)\n",
    "vp.fit(sentiments)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-02-18T13:50:29.375550Z",
     "start_time": "2019-02-18T13:50:20.461066Z"
    }
   },
   "outputs": [],
   "source": [
    "X = np.zeros_like(np.array(sentiments))\n",
    "X = np.array(list(vp.transform(sentiments)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-02-18T13:51:10.813060Z",
     "start_time": "2019-02-18T13:51:10.803114Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(64720, 3000)"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-02-18T13:51:26.524775Z",
     "start_time": "2019-02-18T13:51:23.680466Z"
    }
   },
   "outputs": [],
   "source": [
    "X_train, X_val, y_train, y_val = train_test_split(X, y, test_size=0.4)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-24T17:22:05.056797Z",
     "start_time": "2018-11-24T17:21:54.396466Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "GaussianNB(priors=None)"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "gnb = GaussianNB()\n",
    "gnb.fit(X_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-24T17:22:11.215095Z",
     "start_time": "2018-11-24T17:22:05.073036Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.5172666872682324"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "gnb.score(X_val, y_val)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-02-18T13:51:28.530488Z",
     "start_time": "2019-02-18T13:51:27.817875Z"
    }
   },
   "outputs": [],
   "source": [
    "trainX = pad_sequences(X_train, maxlen=100)\n",
    "valX = pad_sequences(X_val, maxlen=100)\n",
    "trainY = to_categorical(y_train, num_classes=2)\n",
    "valY = to_categorical(y_val, num_classes=2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-02-18T13:51:35.577310Z",
     "start_time": "2019-02-18T13:51:35.567575Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(38832, 100)"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "trainX.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-24T17:58:13.818483Z",
     "start_time": "2018-11-24T17:22:12.353582Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 38832 samples, validate on 25888 samples\n",
      "Epoch 1/5\n",
      "38832/38832 [==============================] - 472s 12ms/step - loss: 0.6937 - acc: 0.5035 - val_loss: 0.6930 - val_acc: 0.5110\n",
      "Epoch 2/5\n",
      "38832/38832 [==============================] - 415s 11ms/step - loss: 0.6935 - acc: 0.5030 - val_loss: 0.6930 - val_acc: 0.5110\n",
      "Epoch 3/5\n",
      "38832/38832 [==============================] - 418s 11ms/step - loss: 0.6934 - acc: 0.5045 - val_loss: 0.6936 - val_acc: 0.4890\n",
      "Epoch 4/5\n",
      "38832/38832 [==============================] - 416s 11ms/step - loss: 0.6934 - acc: 0.5031 - val_loss: 0.6931 - val_acc: 0.5110\n",
      "Epoch 5/5\n",
      "38832/38832 [==============================] - 435s 11ms/step - loss: 0.6932 - acc: 0.5043 - val_loss: 0.6931 - val_acc: 0.5110\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.callbacks.History at 0x1c50ea4358>"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model = Sequential()\n",
    "model.add(Embedding(input_dim=len(sentiments), output_dim=128))\n",
    "model.add(LSTM(128, dropout=0.8))\n",
    "model.add(Dense(2, activation='softmax'))\n",
    "model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])\n",
    "          \n",
    "model.fit(trainX, trainY, validation_data=(valX, valY), batch_size=1000, epochs=5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-24T18:00:36.314660Z",
     "start_time": "2018-11-24T17:58:13.858411Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "25888/25888 [==============================] - 142s 6ms/step\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[0.6930594986076702, 0.5109703315718949]"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.evaluate(valX, valY, batch_size=100)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "RNN具有记忆性，在经过大量训练后可以学习到时序数据的潜在规律，并且可以使用这种规律随机生成新的序列。"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-02-18T11:55:24.459406Z",
     "start_time": "2019-02-18T11:55:19.872684Z"
    }
   },
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "from collections import Counter\n",
    "import matplotlib.pyplot as plt\n",
    "plt.style.use('classic')\n",
    "import warnings\n",
    "import nltk\n",
    "import glob\n",
    "warnings.filterwarnings('ignore')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-02-18T11:55:33.342347Z",
     "start_time": "2019-02-18T11:55:24.464626Z"
    }
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "from keras.models import Sequential\n",
    "from keras.layers.core import Dense, Activation, SpatialDropout1D\n",
    "from keras.layers.embeddings import Embedding\n",
    "from keras.layers.recurrent import LSTM\n",
    "from keras.preprocessing.sequence import pad_sequences\n",
    "from keras.utils import to_categorical\n",
    "from sklearn.model_selection import train_test_split"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-02-18T11:55:33.360073Z",
     "start_time": "2019-02-18T11:55:33.345565Z"
    }
   },
   "outputs": [],
   "source": [
    "FILENAME = './数据集/US_Cities.txt'\n",
    "\n",
    "def build_vocab(filename):\n",
    "    counter = Counter()\n",
    "    f = open(filename, 'r')\n",
    "    for line in f:\n",
    "        for c in '_'.join(line.strip()).replace(' ', 'SPACE').split('_'):\n",
    "            counter[c] += 1\n",
    "    chr2idx = {c:i+3 for i, c in enumerate(counter)}\n",
    "    chr2idx['SOS'] = 0\n",
    "    chr2idx['EOS'] = 1\n",
    "    chr2idx['UNK'] = 2\n",
    "    idx2chr = {v:k for k, v in chr2idx.items()}\n",
    "    return chr2idx, idx2chr"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-02-18T11:55:33.373927Z",
     "start_time": "2019-02-18T11:55:33.366046Z"
    }
   },
   "outputs": [],
   "source": [
    "def get_maxlen(filename):\n",
    "    f = open(filename, 'r')\n",
    "    sent_len = 0\n",
    "    sent_maxlen = 0\n",
    "    for line in f:\n",
    "        sent_len = len(line.strip()) + 2\n",
    "        if sent_len > sent_maxlen:\n",
    "            sent_maxlen = sent_len\n",
    "    return sent_maxlen"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-02-18T11:55:33.390413Z",
     "start_time": "2019-02-18T11:55:33.378912Z"
    }
   },
   "outputs": [],
   "source": [
    "def vectorize(filename, chr2idx, seqlen, step):\n",
    "    f = open(filename, 'r')\n",
    "    text = []\n",
    "    for line in f:\n",
    "        line = ['SOS'] + '_'.join(line.strip()).replace(' ', 'SPACE').split('_') + ['EOS']\n",
    "        for word in line:\n",
    "            text.append(word)\n",
    "    input_chars = []\n",
    "    label_chars = []\n",
    "    for i in range(0, len(text) - seqlen, step):\n",
    "        input_chars.append(text[i:i + seqlen])\n",
    "        label_chars.append(text[i + seqlen])\n",
    "    X = np.zeros((len(input_chars), seqlen, len(chr2idx)))\n",
    "    y = np.zeros((len(input_chars), len(chr2idx)))\n",
    "    for i, input_char in enumerate(input_chars):\n",
    "        for j, char in enumerate(input_char):\n",
    "            X[i, j, chr2idx[char]] = 1\n",
    "        y[i, chr2idx[label_chars[i]]] = 1\n",
    "    return X, y"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-02-18T11:55:52.419474Z",
     "start_time": "2019-02-18T11:55:33.395013Z"
    }
   },
   "outputs": [],
   "source": [
    "STEP = 1\n",
    "SEQLEN = get_maxlen(FILENAME)\n",
    "\n",
    "chr2idx, idx2chr = build_vocab(FILENAME)\n",
    "X, y = vectorize(FILENAME, chr2idx, SEQLEN, STEP)\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-24T19:47:32.126444Z",
     "start_time": "2018-11-24T18:01:10.268781Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 182071 samples, validate on 45518 samples\n",
      "Epoch 1/10\n",
      "182071/182071 [==============================] - 702s 4ms/step - loss: 2.6196 - acc: 0.2818 - val_loss: 2.1920 - val_acc: 0.3590\n",
      "Epoch 2/10\n",
      "182071/182071 [==============================] - 635s 3ms/step - loss: 2.0403 - acc: 0.4051 - val_loss: 1.8658 - val_acc: 0.4716\n",
      "Epoch 3/10\n",
      "182071/182071 [==============================] - 639s 4ms/step - loss: 1.6745 - acc: 0.5336 - val_loss: 1.5140 - val_acc: 0.5817\n",
      "Epoch 4/10\n",
      "182071/182071 [==============================] - 640s 4ms/step - loss: 1.4124 - acc: 0.6096 - val_loss: 1.3553 - val_acc: 0.6262\n",
      "Epoch 5/10\n",
      "182071/182071 [==============================] - 612s 3ms/step - loss: 1.2881 - acc: 0.6432 - val_loss: 1.2674 - val_acc: 0.6503\n",
      "Epoch 6/10\n",
      "182071/182071 [==============================] - 625s 3ms/step - loss: 1.2125 - acc: 0.6626 - val_loss: 1.2112 - val_acc: 0.6632\n",
      "Epoch 7/10\n",
      "182071/182071 [==============================] - 631s 3ms/step - loss: 1.1602 - acc: 0.6753 - val_loss: 1.1758 - val_acc: 0.6715\n",
      "Epoch 8/10\n",
      "182071/182071 [==============================] - 624s 3ms/step - loss: 1.1199 - acc: 0.6847 - val_loss: 1.1492 - val_acc: 0.6790\n",
      "Epoch 9/10\n",
      "182071/182071 [==============================] - 635s 3ms/step - loss: 1.0869 - acc: 0.6934 - val_loss: 1.1270 - val_acc: 0.6830\n",
      "Epoch 10/10\n",
      "182071/182071 [==============================] - 616s 3ms/step - loss: 1.0597 - acc: 0.6996 - val_loss: 1.1071 - val_acc: 0.6904\n"
     ]
    }
   ],
   "source": [
    "HIDDEN_SIZE = 128\n",
    "BATCH_SIZE = 256\n",
    "NUM_EPOCHS = 10\n",
    "\n",
    "model = Sequential()\n",
    "model.add(LSTM(HIDDEN_SIZE, return_sequences=False, input_shape=(SEQLEN, len(chr2idx)), unroll=True))\n",
    "model.add(Dense(len(chr2idx)))\n",
    "model.add(Activation('softmax'))\n",
    "model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])\n",
    "history = model.fit(X_train, y_train, batch_size=BATCH_SIZE, epochs=NUM_EPOCHS, validation_data=(X_test, y_test))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-24T19:48:32.694468Z",
     "start_time": "2018-11-24T19:47:32.152432Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "45518/45518 [==============================] - 60s 1ms/step\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[1.1071459434222994, 0.690430159534007]"
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "model.evaluate(X_test, y_test, verbose=1, batch_size=BATCH_SIZE)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-24T19:50:35.591878Z",
     "start_time": "2018-11-24T19:48:32.698376Z"
    }
   },
   "outputs": [],
   "source": [
    "y_pred = model.predict(X_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-24T19:50:35.617629Z",
     "start_time": "2018-11-24T19:50:35.596637Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['EOS o i r SOS EOS C SOS r M e n f e EOS w SOS h O o SPACE EOS y r t EOS t SOS t i y p e R i c c i o n i e n a EOS e SOS SOS y a n a g r l G o y EOS l n v EOS K e EOS o a o a a R e EOS e M SOS o a e a o n n K s l o N r d e EOS EOS SPACE o r l r SOS']\n"
     ]
    }
   ],
   "source": [
    "NUM_DISPLAY = 100\n",
    "\n",
    "sentence = []\n",
    "sentences = []\n",
    "idx = []\n",
    "for i in range(NUM_DISPLAY):\n",
    "    idx.append(np.argmax(y_pred[i]))\n",
    "for j in idx:\n",
    "    sentence.append(idx2chr[j])\n",
    "sentences.append(' '.join(sentence))\n",
    "print(sentences)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-24T19:50:36.700633Z",
     "start_time": "2018-11-24T19:50:35.625949Z"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x1c786bd7b8>"
      ]
     },
     "execution_count": 39,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.subplot(211)\n",
    "plt.title(\"Accuracy\")\n",
    "plt.plot(history.history[\"acc\"], color=\"g\", label=\"train\")\n",
    "plt.plot(history.history[\"val_acc\"], color=\"b\", label=\"validation\")\n",
    "plt.legend(loc=\"best\")\n",
    "\n",
    "plt.subplot(212)\n",
    "plt.title(\"Loss\")\n",
    "plt.plot(history.history[\"loss\"], color=\"g\", label=\"train\")\n",
    "plt.plot(history.history[\"val_loss\"], color=\"b\", label=\"validation\")\n",
    "plt.legend(loc=\"best\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-24T23:03:06.367556Z",
     "start_time": "2018-11-24T19:50:36.714480Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "==================================================\n",
      "Iteration #: 0\n",
      "Epoch 1/1\n",
      "227589/227589 [==============================] - 722s 3ms/step - loss: 1.0468 - acc: 0.7031\n",
      "Generating from seed: -\n",
      "-e\n",
      "==================================================\n",
      "Iteration #: 1\n",
      "Epoch 1/1\n",
      "227589/227589 [==============================] - 700s 3ms/step - loss: 1.0226 - acc: 0.7084\n",
      "Generating from seed: e\n",
      "-ee\n",
      "==================================================\n",
      "Iteration #: 2\n",
      "Epoch 1/1\n",
      "227589/227589 [==============================] - 728s 3ms/step - loss: 1.0032 - acc: 0.7134\n",
      "Generating from seed: e\n",
      "-eee\n",
      "==================================================\n",
      "Iteration #: 3\n",
      "Epoch 1/1\n",
      "227589/227589 [==============================] - 701s 3ms/step - loss: 0.9820 - acc: 0.7178\n",
      "Generating from seed: e\n",
      "-eeee\n",
      "==================================================\n",
      "Iteration #: 4\n",
      "Epoch 1/1\n",
      "227589/227589 [==============================] - 717s 3ms/step - loss: 0.9641 - acc: 0.7223\n",
      "Generating from seed: e\n",
      "-eeeeo\n",
      "==================================================\n",
      "Iteration #: 5\n",
      "Epoch 1/1\n",
      "227589/227589 [==============================] - 703s 3ms/step - loss: 0.9482 - acc: 0.7256\n",
      "Generating from seed: o\n",
      "-eeeeoo\n",
      "==================================================\n",
      "Iteration #: 6\n",
      "Epoch 1/1\n",
      "227589/227589 [==============================] - 525s 2ms/step - loss: 0.9334 - acc: 0.7296\n",
      "Generating from seed: o\n",
      "-eeeeooo\n",
      "==================================================\n",
      "Iteration #: 7\n",
      "Epoch 1/1\n",
      "227589/227589 [==============================] - 269s 1ms/step - loss: 0.9213 - acc: 0.7322\n",
      "Generating from seed: o\n",
      "-eeeeoooo\n",
      "==================================================\n",
      "Iteration #: 8\n",
      "Epoch 1/1\n",
      "227589/227589 [==============================] - 269s 1ms/step - loss: 0.9056 - acc: 0.7352\n",
      "Generating from seed: o\n",
      "-eeeeooooo\n",
      "==================================================\n",
      "Iteration #: 9\n",
      "Epoch 1/1\n",
      "227589/227589 [==============================] - 272s 1ms/step - loss: 0.8943 - acc: 0.7384\n",
      "Generating from seed: o\n",
      "-eeeeoooooo\n",
      "==================================================\n",
      "Iteration #: 10\n",
      "Epoch 1/1\n",
      "227589/227589 [==============================] - 274s 1ms/step - loss: 0.8827 - acc: 0.7410\n",
      "Generating from seed: o\n",
      "-eeeeooooooo\n",
      "==================================================\n",
      "Iteration #: 11\n",
      "Epoch 1/1\n",
      "227589/227589 [==============================] - 270s 1ms/step - loss: 0.8714 - acc: 0.7433\n",
      "Generating from seed: o\n",
      "-eeeeoooooooo\n",
      "==================================================\n",
      "Iteration #: 12\n",
      "Epoch 1/1\n",
      "227589/227589 [==============================] - 270s 1ms/step - loss: 0.8611 - acc: 0.7463\n",
      "Generating from seed: o\n",
      "-eeeeooooooooo\n",
      "==================================================\n",
      "Iteration #: 13\n",
      "Epoch 1/1\n",
      "227589/227589 [==============================] - 272s 1ms/step - loss: 0.8499 - acc: 0.7488\n",
      "Generating from seed: o\n",
      "-eeeeoooooooooo\n",
      "==================================================\n",
      "Iteration #: 14\n",
      "Epoch 1/1\n",
      "227589/227589 [==============================] - 270s 1ms/step - loss: 0.8400 - acc: 0.7513\n",
      "Generating from seed: o\n",
      "-eeeeooooooooooo\n",
      "==================================================\n",
      "Iteration #: 15\n",
      "Epoch 1/1\n",
      "227589/227589 [==============================] - 269s 1ms/step - loss: 0.8308 - acc: 0.7539\n",
      "Generating from seed: o\n",
      "-eeeeoooooooooooo\n",
      "==================================================\n",
      "Iteration #: 16\n",
      "Epoch 1/1\n",
      "227589/227589 [==============================] - 270s 1ms/step - loss: 0.8208 - acc: 0.7563\n",
      "Generating from seed: o\n",
      "-eeeeooooooooooooo\n",
      "==================================================\n",
      "Iteration #: 17\n",
      "Epoch 1/1\n",
      "227589/227589 [==============================] - 270s 1ms/step - loss: 0.8118 - acc: 0.7592\n",
      "Generating from seed: o\n",
      "-eeeeoooooooooooooEOS\n",
      "==================================================\n",
      "Iteration #: 18\n",
      "Epoch 1/1\n",
      "227589/227589 [==============================] - 269s 1ms/step - loss: 0.8058 - acc: 0.7601\n",
      "Generating from seed: EOS\n",
      "-eeeeoooooooooooooEOSEOS\n",
      "==================================================\n",
      "Iteration #: 19\n",
      "Epoch 1/1\n",
      "227589/227589 [==============================] - 269s 1ms/step - loss: 0.7949 - acc: 0.7631\n",
      "Generating from seed: EOS\n",
      "-eeeeoooooooooooooEOSEOSu\n",
      "==================================================\n",
      "Iteration #: 20\n",
      "Epoch 1/1\n",
      "227589/227589 [==============================] - 269s 1ms/step - loss: 0.7867 - acc: 0.7651\n",
      "Generating from seed: u\n",
      "-eeeeoooooooooooooEOSEOSuy\n",
      "==================================================\n",
      "Iteration #: 21\n",
      "Epoch 1/1\n",
      "227589/227589 [==============================] - 269s 1ms/step - loss: 0.7773 - acc: 0.7677\n",
      "Generating from seed: y\n",
      "-eeeeoooooooooooooEOSEOSuyg\n",
      "==================================================\n",
      "Iteration #: 22\n",
      "Epoch 1/1\n",
      "227589/227589 [==============================] - 269s 1ms/step - loss: 0.7693 - acc: 0.7696\n",
      "Generating from seed: g\n",
      "-eeeeoooooooooooooEOSEOSuygo\n",
      "==================================================\n",
      "Iteration #: 23\n",
      "Epoch 1/1\n",
      "227589/227589 [==============================] - 273s 1ms/step - loss: 0.7630 - acc: 0.7707\n",
      "Generating from seed: o\n",
      "-eeeeoooooooooooooEOSEOSuygoo\n",
      "==================================================\n",
      "Iteration #: 24\n",
      "Epoch 1/1\n",
      "227589/227589 [==============================] - 272s 1ms/step - loss: 0.7552 - acc: 0.7730\n",
      "Generating from seed: o\n",
      "-eeeeoooooooooooooEOSEOSuygooo\n",
      "==================================================\n",
      "Iteration #: 25\n",
      "Epoch 1/1\n",
      "227589/227589 [==============================] - 269s 1ms/step - loss: 0.7459 - acc: 0.7760\n",
      "Generating from seed: o\n",
      "-eeeeoooooooooooooEOSEOSuygoooo\n",
      "==================================================\n",
      "Iteration #: 26\n",
      "Epoch 1/1\n",
      "227589/227589 [==============================] - 269s 1ms/step - loss: 0.7411 - acc: 0.7776\n",
      "Generating from seed: o\n",
      "-eeeeoooooooooooooEOSEOSuygooool\n",
      "==================================================\n",
      "Iteration #: 27\n",
      "Epoch 1/1\n",
      "227589/227589 [==============================] - 269s 1ms/step - loss: 0.7325 - acc: 0.7790\n",
      "Generating from seed: l\n",
      "-eeeeoooooooooooooEOSEOSuygoooolo\n",
      "==================================================\n",
      "Iteration #: 28\n",
      "Epoch 1/1\n",
      "227589/227589 [==============================] - 270s 1ms/step - loss: 0.7248 - acc: 0.7817\n",
      "Generating from seed: o\n",
      "-eeeeoooooooooooooEOSEOSuygooooloy\n",
      "==================================================\n",
      "Iteration #: 29\n",
      "Epoch 1/1\n",
      "227589/227589 [==============================] - 270s 1ms/step - loss: 0.7197 - acc: 0.7830\n",
      "Generating from seed: y\n",
      "-eeeeoooooooooooooEOSEOSuygooooloyEOS\n",
      "==================================================\n",
      "Iteration #: 30\n",
      "Epoch 1/1\n",
      "227589/227589 [==============================] - 270s 1ms/step - loss: 0.7121 - acc: 0.7851\n",
      "Generating from seed: EOS\n",
      "-eeeeoooooooooooooEOSEOSuygooooloyEOSp\n",
      "==================================================\n",
      "Iteration #: 31\n",
      "Epoch 1/1\n",
      "227589/227589 [==============================] - 270s 1ms/step - loss: 0.7079 - acc: 0.7870\n",
      "Generating from seed: p\n",
      "-eeeeoooooooooooooEOSEOSuygooooloyEOSpEOS\n"
     ]
    }
   ],
   "source": [
    "NUM_ITERATIONS = 32\n",
    "NUM_EPOCHS_PER_ITERATION = 1\n",
    "NUM_PREDS_PER_EPOCH = 10\n",
    "\n",
    "sentence = ''\n",
    "test_idx = np.random.randint(len(chr2idx))\n",
    "test_chars = seed = idx2chr[test_idx]\n",
    "for iteration in range(NUM_ITERATIONS):\n",
    "    print(\"=\" * 50)\n",
    "    print(\"Iteration #: %d\" %(iteration))\n",
    "    model.fit(X, y, batch_size=BATCH_SIZE, epochs=NUM_EPOCHS_PER_ITERATION)\n",
    "    print(\"Generating from seed: %s\" %(test_chars))\n",
    "    for i in range(NUM_EPOCHS_PER_ITERATION):\n",
    "        X_test = np.zeros((1, SEQLEN, len(chr2idx)))\n",
    "        for j, ch in enumerate(test_chars):\n",
    "            X_test[0, j, chr2idx[ch]] = 1\n",
    "        pred = model.predict(X_test, verbose=0)[0]\n",
    "        ypred = idx2chr[np.argmax(pred)]\n",
    "        sentence += ypred\n",
    "        test_chars = ypred\n",
    "    print(seed + sentence)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-24T23:03:33.384308Z",
     "start_time": "2018-11-24T23:03:06.383448Z"
    }
   },
   "outputs": [],
   "source": [
    "FILENAME = './数据集/Password/wvs-pass.txt'\n",
    "\n",
    "STEP = 1\n",
    "SEQLEN = get_maxlen(FILENAME)\n",
    "\n",
    "chr2idx, idx2chr = build_vocab(FILENAME)\n",
    "X, y = vectorize(FILENAME, chr2idx, SEQLEN, STEP)\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-24T23:51:22.879020Z",
     "start_time": "2018-11-24T23:03:33.398339Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 153650 samples, validate on 38413 samples\n",
      "Epoch 1/10\n",
      "153650/153650 [==============================] - 327s 2ms/step - loss: 3.0754 - acc: 0.2522 - val_loss: 2.5881 - val_acc: 0.3402\n",
      "Epoch 2/10\n",
      "153650/153650 [==============================] - 280s 2ms/step - loss: 2.4535 - acc: 0.3675 - val_loss: 2.3638 - val_acc: 0.3813\n",
      "Epoch 3/10\n",
      "153650/153650 [==============================] - 280s 2ms/step - loss: 2.2971 - acc: 0.3984 - val_loss: 2.2486 - val_acc: 0.4097\n",
      "Epoch 4/10\n",
      "153650/153650 [==============================] - 280s 2ms/step - loss: 2.1901 - acc: 0.4248 - val_loss: 2.1546 - val_acc: 0.4360\n",
      "Epoch 5/10\n",
      "153650/153650 [==============================] - 285s 2ms/step - loss: 2.0987 - acc: 0.4508 - val_loss: 2.0834 - val_acc: 0.4541\n",
      "Epoch 6/10\n",
      "153650/153650 [==============================] - 281s 2ms/step - loss: 2.0205 - acc: 0.4717 - val_loss: 2.0221 - val_acc: 0.4698\n",
      "Epoch 7/10\n",
      "153650/153650 [==============================] - 280s 2ms/step - loss: 1.9540 - acc: 0.4882 - val_loss: 1.9714 - val_acc: 0.4869\n",
      "Epoch 8/10\n",
      "153650/153650 [==============================] - 281s 2ms/step - loss: 1.8951 - acc: 0.5035 - val_loss: 1.9166 - val_acc: 0.5010\n",
      "Epoch 9/10\n",
      "153650/153650 [==============================] - 280s 2ms/step - loss: 1.8416 - acc: 0.5167 - val_loss: 1.8793 - val_acc: 0.5119\n",
      "Epoch 10/10\n",
      "153650/153650 [==============================] - 281s 2ms/step - loss: 1.7956 - acc: 0.5282 - val_loss: 1.8481 - val_acc: 0.5189\n"
     ]
    }
   ],
   "source": [
    "HIDDEN_SIZE = 128\n",
    "BATCH_SIZE = 256\n",
    "NUM_EPOCHS = 10\n",
    "\n",
    "model = Sequential()\n",
    "model.add(LSTM(HIDDEN_SIZE, return_sequences=False, input_shape=(SEQLEN, len(chr2idx)), unroll=True))\n",
    "model.add(Dense(len(chr2idx)))\n",
    "model.add(Activation('softmax'))\n",
    "model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])\n",
    "history = model.fit(X_train, y_train, batch_size=BATCH_SIZE, epochs=NUM_EPOCHS, validation_data=(X_test, y_test))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-25T10:56:43.106598Z",
     "start_time": "2018-11-24T23:51:22.903269Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "==================================================\n",
      "Iteration #: 0\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 331s 2ms/step - loss: 1.7653 - acc: 0.5364\n",
      "Generating from seed: SOS\n",
      "SOS·\n",
      "==================================================\n",
      "Iteration #: 1\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 318s 2ms/step - loss: 1.7197 - acc: 0.5477\n",
      "Generating from seed: ·\n",
      "SOS·y\n",
      "==================================================\n",
      "Iteration #: 2\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 317s 2ms/step - loss: 1.6778 - acc: 0.5582\n",
      "Generating from seed: y\n",
      "SOS·yy\n",
      "==================================================\n",
      "Iteration #: 3\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 317s 2ms/step - loss: 1.6400 - acc: 0.5677\n",
      "Generating from seed: y\n",
      "SOS·yyy\n",
      "==================================================\n",
      "Iteration #: 4\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 317s 2ms/step - loss: 1.6039 - acc: 0.5767\n",
      "Generating from seed: y\n",
      "SOS·yyy1\n",
      "==================================================\n",
      "Iteration #: 5\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 318s 2ms/step - loss: 1.5710 - acc: 0.5850\n",
      "Generating from seed: 1\n",
      "SOS·yyy1s\n",
      "==================================================\n",
      "Iteration #: 6\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 321s 2ms/step - loss: 1.5407 - acc: 0.5925\n",
      "Generating from seed: s\n",
      "SOS·yyy1sd\n",
      "==================================================\n",
      "Iteration #: 7\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 320s 2ms/step - loss: 1.5118 - acc: 0.5998\n",
      "Generating from seed: d\n",
      "SOS·yyy1sdd\n",
      "==================================================\n",
      "Iteration #: 8\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 318s 2ms/step - loss: 1.4844 - acc: 0.6076\n",
      "Generating from seed: d\n",
      "SOS·yyy1sddd\n",
      "==================================================\n",
      "Iteration #: 9\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 318s 2ms/step - loss: 1.4585 - acc: 0.6135\n",
      "Generating from seed: d\n",
      "SOS·yyy1sdddEOS\n",
      "==================================================\n",
      "Iteration #: 10\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 318s 2ms/step - loss: 1.4329 - acc: 0.6202\n",
      "Generating from seed: EOS\n",
      "SOS·yyy1sdddEOSf\n",
      "==================================================\n",
      "Iteration #: 11\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 318s 2ms/step - loss: 1.4101 - acc: 0.6256\n",
      "Generating from seed: f\n",
      "SOS·yyy1sdddEOSf-\n",
      "==================================================\n",
      "Iteration #: 12\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 319s 2ms/step - loss: 1.3880 - acc: 0.6317\n",
      "Generating from seed: -\n",
      "SOS·yyy1sdddEOSf-s\n",
      "==================================================\n",
      "Iteration #: 13\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 318s 2ms/step - loss: 1.3664 - acc: 0.6362\n",
      "Generating from seed: s\n",
      "SOS·yyy1sdddEOSf-sd\n",
      "==================================================\n",
      "Iteration #: 14\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 319s 2ms/step - loss: 1.3453 - acc: 0.6420\n",
      "Generating from seed: d\n",
      "SOS·yyy1sdddEOSf-sdh\n",
      "==================================================\n",
      "Iteration #: 15\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 319s 2ms/step - loss: 1.3266 - acc: 0.6465\n",
      "Generating from seed: h\n",
      "SOS·yyy1sdddEOSf-sdhd\n",
      "==================================================\n",
      "Iteration #: 16\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 318s 2ms/step - loss: 1.3083 - acc: 0.6513\n",
      "Generating from seed: d\n",
      "SOS·yyy1sdddEOSf-sdhdf\n",
      "==================================================\n",
      "Iteration #: 17\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 321s 2ms/step - loss: 1.2913 - acc: 0.6552\n",
      "Generating from seed: f\n",
      "SOS·yyy1sdddEOSf-sdhdfq\n",
      "==================================================\n",
      "Iteration #: 18\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 320s 2ms/step - loss: 1.2731 - acc: 0.6602\n",
      "Generating from seed: q\n",
      "SOS·yyy1sdddEOSf-sdhdfqL\n",
      "==================================================\n",
      "Iteration #: 19\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 320s 2ms/step - loss: 1.2577 - acc: 0.6643\n",
      "Generating from seed: L\n",
      "SOS·yyy1sdddEOSf-sdhdfqLs\n",
      "==================================================\n",
      "Iteration #: 20\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 328s 2ms/step - loss: 1.2415 - acc: 0.6678\n",
      "Generating from seed: s\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsy\n",
      "==================================================\n",
      "Iteration #: 21\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 318s 2ms/step - loss: 1.2270 - acc: 0.6721\n",
      "Generating from seed: y\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyh\n",
      "==================================================\n",
      "Iteration #: 22\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 319s 2ms/step - loss: 1.2121 - acc: 0.6764\n",
      "Generating from seed: h\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS\n",
      "==================================================\n",
      "Iteration #: 23\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 369s 2ms/step - loss: 1.1983 - acc: 0.6789\n",
      "Generating from seed: EOS\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS7\n",
      "==================================================\n",
      "Iteration #: 24\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 722s 4ms/step - loss: 1.1870 - acc: 0.6825\n",
      "Generating from seed: 7\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS75\n",
      "==================================================\n",
      "Iteration #: 25\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 780s 4ms/step - loss: 1.1727 - acc: 0.6867\n",
      "Generating from seed: 5\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755\n",
      "==================================================\n",
      "Iteration #: 26\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 699s 4ms/step - loss: 1.1593 - acc: 0.6900\n",
      "Generating from seed: 5\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755d\n",
      "==================================================\n",
      "Iteration #: 27\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 368s 2ms/step - loss: 1.1494 - acc: 0.6926\n",
      "Generating from seed: d\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dy\n",
      "==================================================\n",
      "Iteration #: 28\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 359s 2ms/step - loss: 1.1382 - acc: 0.6950\n",
      "Generating from seed: y\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyf\n",
      "==================================================\n",
      "Iteration #: 29\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 741s 4ms/step - loss: 1.1258 - acc: 0.6990\n",
      "Generating from seed: f\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyft\n",
      "==================================================\n",
      "Iteration #: 30\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 882s 5ms/step - loss: 1.1161 - acc: 0.7010\n",
      "Generating from seed: t\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth\n",
      "==================================================\n",
      "Iteration #: 31\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 951s 5ms/step - loss: 1.1064 - acc: 0.7041\n",
      "Generating from seed: h\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth5\n",
      "==================================================\n",
      "Iteration #: 32\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 916s 5ms/step - loss: 1.0966 - acc: 0.7056\n",
      "Generating from seed: 5\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55\n",
      "==================================================\n",
      "Iteration #: 33\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 864s 4ms/step - loss: 1.0877 - acc: 0.7097\n",
      "Generating from seed: 5\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth555\n",
      "==================================================\n",
      "Iteration #: 34\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 978s 5ms/step - loss: 1.0790 - acc: 0.7113\n",
      "Generating from seed: 5\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth5555\n",
      "==================================================\n",
      "Iteration #: 35\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 734s 4ms/step - loss: 1.0686 - acc: 0.7143\n",
      "Generating from seed: 5\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555\n",
      "==================================================\n",
      "Iteration #: 36\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 471s 2ms/step - loss: 1.0634 - acc: 0.7162\n",
      "Generating from seed: 5\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth555555\n",
      "==================================================\n",
      "Iteration #: 37\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 872s 5ms/step - loss: 1.0527 - acc: 0.7179\n",
      "Generating from seed: 5\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth5555553\n",
      "==================================================\n",
      "Iteration #: 38\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 817s 4ms/step - loss: 1.0463 - acc: 0.7204\n",
      "Generating from seed: 3\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535\n",
      "==================================================\n",
      "Iteration #: 39\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 580s 3ms/step - loss: 1.0369 - acc: 0.7226\n",
      "Generating from seed: 5\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth555555353\n",
      "==================================================\n",
      "Iteration #: 40\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 321s 2ms/step - loss: 1.0309 - acc: 0.7243\n",
      "Generating from seed: 3\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth5555553535\n",
      "==================================================\n",
      "Iteration #: 41\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 323s 2ms/step - loss: 1.0241 - acc: 0.7260\n",
      "Generating from seed: 5\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357\n",
      "==================================================\n",
      "Iteration #: 42\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 321s 2ms/step - loss: 1.0188 - acc: 0.7268\n",
      "Generating from seed: 7\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q\n",
      "==================================================\n",
      "Iteration #: 43\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 321s 2ms/step - loss: 1.0099 - acc: 0.7295\n",
      "Generating from seed: Q\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q7\n",
      "==================================================\n",
      "Iteration #: 44\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 319s 2ms/step - loss: 1.0070 - acc: 0.7302\n",
      "Generating from seed: 7\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q7I\n",
      "==================================================\n",
      "Iteration #: 45\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 321s 2ms/step - loss: 1.0020 - acc: 0.7312\n",
      "Generating from seed: I\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q7IEOS\n",
      "==================================================\n",
      "Iteration #: 46\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 319s 2ms/step - loss: 0.9928 - acc: 0.7344\n",
      "Generating from seed: EOS\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q7IEOSb\n",
      "==================================================\n",
      "Iteration #: 47\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 320s 2ms/step - loss: 0.9884 - acc: 0.7358\n",
      "Generating from seed: b\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q7IEOSbw\n",
      "==================================================\n",
      "Iteration #: 48\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 319s 2ms/step - loss: 0.9830 - acc: 0.7359\n",
      "Generating from seed: w\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q7IEOSbw5\n",
      "==================================================\n",
      "Iteration #: 49\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 319s 2ms/step - loss: 0.9767 - acc: 0.7381\n",
      "Generating from seed: 5\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q7IEOSbw5EOS\n",
      "==================================================\n",
      "Iteration #: 50\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 318s 2ms/step - loss: 0.9732 - acc: 0.7390\n",
      "Generating from seed: EOS\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q7IEOSbw5EOSq\n",
      "==================================================\n",
      "Iteration #: 51\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 319s 2ms/step - loss: 0.9724 - acc: 0.7388\n",
      "Generating from seed: q\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q7IEOSbw5EOSqEOS\n",
      "==================================================\n",
      "Iteration #: 52\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 324s 2ms/step - loss: 0.9662 - acc: 0.7408\n",
      "Generating from seed: EOS\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q7IEOSbw5EOSqEOS5\n",
      "==================================================\n",
      "Iteration #: 53\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 320s 2ms/step - loss: 0.9587 - acc: 0.7424\n",
      "Generating from seed: 5\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q7IEOSbw5EOSqEOS58\n",
      "==================================================\n",
      "Iteration #: 54\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 319s 2ms/step - loss: 0.9548 - acc: 0.7438\n",
      "Generating from seed: 8\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q7IEOSbw5EOSqEOS589\n",
      "==================================================\n",
      "Iteration #: 55\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 321s 2ms/step - loss: 0.9521 - acc: 0.7442\n",
      "Generating from seed: 9\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q7IEOSbw5EOSqEOS589f\n",
      "==================================================\n",
      "Iteration #: 56\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 319s 2ms/step - loss: 0.9488 - acc: 0.7455\n",
      "Generating from seed: f\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q7IEOSbw5EOSqEOS589fq\n",
      "==================================================\n",
      "Iteration #: 57\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 318s 2ms/step - loss: 0.9397 - acc: 0.7485\n",
      "Generating from seed: q\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q7IEOSbw5EOSqEOS589fqh\n",
      "==================================================\n",
      "Iteration #: 58\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 320s 2ms/step - loss: 0.9393 - acc: 0.7472\n",
      "Generating from seed: h\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q7IEOSbw5EOSqEOS589fqh^\n",
      "==================================================\n",
      "Iteration #: 59\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 320s 2ms/step - loss: 0.9345 - acc: 0.7486\n",
      "Generating from seed: ^\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q7IEOSbw5EOSqEOS589fqh^z\n",
      "==================================================\n",
      "Iteration #: 60\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 319s 2ms/step - loss: 0.9304 - acc: 0.7505\n",
      "Generating from seed: z\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q7IEOSbw5EOSqEOS589fqh^zy\n",
      "==================================================\n",
      "Iteration #: 61\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 320s 2ms/step - loss: 0.9289 - acc: 0.7503\n",
      "Generating from seed: y\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q7IEOSbw5EOSqEOS589fqh^zyEOS\n",
      "==================================================\n",
      "Iteration #: 62\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 319s 2ms/step - loss: 0.9247 - acc: 0.7514\n",
      "Generating from seed: EOS\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q7IEOSbw5EOSqEOS589fqh^zyEOS0\n",
      "==================================================\n",
      "Iteration #: 63\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 319s 2ms/step - loss: 0.9239 - acc: 0.7520\n",
      "Generating from seed: 0\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q7IEOSbw5EOSqEOS589fqh^zyEOS0A\n",
      "==================================================\n",
      "Iteration #: 64\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 325s 2ms/step - loss: 0.9177 - acc: 0.7525\n",
      "Generating from seed: A\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q7IEOSbw5EOSqEOS589fqh^zyEOS0AEOS\n",
      "==================================================\n",
      "Iteration #: 65\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 350s 2ms/step - loss: 0.9141 - acc: 0.7540\n",
      "Generating from seed: EOS\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q7IEOSbw5EOSqEOS589fqh^zyEOS0AEOSz\n",
      "==================================================\n",
      "Iteration #: 66\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 416s 2ms/step - loss: 0.9130 - acc: 0.7541\n",
      "Generating from seed: z\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q7IEOSbw5EOSqEOS589fqh^zyEOS0AEOSzz\n",
      "==================================================\n",
      "Iteration #: 67\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 421s 2ms/step - loss: 0.9114 - acc: 0.7545\n",
      "Generating from seed: z\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q7IEOSbw5EOSqEOS589fqh^zyEOS0AEOSzz$\n",
      "==================================================\n",
      "Iteration #: 68\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 349s 2ms/step - loss: 0.9082 - acc: 0.7553\n",
      "Generating from seed: $\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q7IEOSbw5EOSqEOS589fqh^zyEOS0AEOSzz$5\n",
      "==================================================\n",
      "Iteration #: 69\n",
      "Epoch 1/1\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "192063/192063 [==============================] - 352s 2ms/step - loss: 0.9051 - acc: 0.7558\n",
      "Generating from seed: 5\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q7IEOSbw5EOSqEOS589fqh^zyEOS0AEOSzz$5EOS\n",
      "==================================================\n",
      "Iteration #: 70\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 347s 2ms/step - loss: 0.9003 - acc: 0.7567\n",
      "Generating from seed: EOS\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q7IEOSbw5EOSqEOS589fqh^zyEOS0AEOSzz$5EOS4\n",
      "==================================================\n",
      "Iteration #: 71\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 349s 2ms/step - loss: 0.8987 - acc: 0.7585\n",
      "Generating from seed: 4\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q7IEOSbw5EOSqEOS589fqh^zyEOS0AEOSzz$5EOS4EOS\n",
      "==================================================\n",
      "Iteration #: 72\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 352s 2ms/step - loss: 0.8949 - acc: 0.7594\n",
      "Generating from seed: EOS\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q7IEOSbw5EOSqEOS589fqh^zyEOS0AEOSzz$5EOS4EOSEOS\n",
      "==================================================\n",
      "Iteration #: 73\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 349s 2ms/step - loss: 0.8935 - acc: 0.7588\n",
      "Generating from seed: EOS\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q7IEOSbw5EOSqEOS589fqh^zyEOS0AEOSzz$5EOS4EOSEOS0\n",
      "==================================================\n",
      "Iteration #: 74\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 363s 2ms/step - loss: 0.8942 - acc: 0.7585\n",
      "Generating from seed: 0\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q7IEOSbw5EOSqEOS589fqh^zyEOS0AEOSzz$5EOS4EOSEOS05\n",
      "==================================================\n",
      "Iteration #: 75\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 362s 2ms/step - loss: 0.8860 - acc: 0.7606\n",
      "Generating from seed: 5\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q7IEOSbw5EOSqEOS589fqh^zyEOS0AEOSzz$5EOS4EOSEOS05u\n",
      "==================================================\n",
      "Iteration #: 76\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 359s 2ms/step - loss: 0.8820 - acc: 0.7631\n",
      "Generating from seed: u\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q7IEOSbw5EOSqEOS589fqh^zyEOS0AEOSzz$5EOS4EOSEOS05uT\n",
      "==================================================\n",
      "Iteration #: 77\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 359s 2ms/step - loss: 0.8827 - acc: 0.7622\n",
      "Generating from seed: T\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q7IEOSbw5EOSqEOS589fqh^zyEOS0AEOSzz$5EOS4EOSEOS05uTs\n",
      "==================================================\n",
      "Iteration #: 78\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 373s 2ms/step - loss: 0.8820 - acc: 0.7624\n",
      "Generating from seed: s\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q7IEOSbw5EOSqEOS589fqh^zyEOS0AEOSzz$5EOS4EOSEOS05uTsEOS\n",
      "==================================================\n",
      "Iteration #: 79\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 365s 2ms/step - loss: 0.8810 - acc: 0.7620\n",
      "Generating from seed: EOS\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q7IEOSbw5EOSqEOS589fqh^zyEOS0AEOSzz$5EOS4EOSEOS05uTsEOSF\n",
      "==================================================\n",
      "Iteration #: 80\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 347s 2ms/step - loss: 0.8795 - acc: 0.7625\n",
      "Generating from seed: F\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q7IEOSbw5EOSqEOS589fqh^zyEOS0AEOSzz$5EOS4EOSEOS05uTsEOSFi\n",
      "==================================================\n",
      "Iteration #: 81\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 354s 2ms/step - loss: 0.8772 - acc: 0.7631\n",
      "Generating from seed: i\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q7IEOSbw5EOSqEOS589fqh^zyEOS0AEOSzz$5EOS4EOSEOS05uTsEOSFiEOS\n",
      "==================================================\n",
      "Iteration #: 82\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 353s 2ms/step - loss: 0.8694 - acc: 0.7653\n",
      "Generating from seed: EOS\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q7IEOSbw5EOSqEOS589fqh^zyEOS0AEOSzz$5EOS4EOSEOS05uTsEOSFiEOSEOS\n",
      "==================================================\n",
      "Iteration #: 83\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 365s 2ms/step - loss: 0.8690 - acc: 0.7651\n",
      "Generating from seed: EOS\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q7IEOSbw5EOSqEOS589fqh^zyEOS0AEOSzz$5EOS4EOSEOS05uTsEOSFiEOSEOSp\n",
      "==================================================\n",
      "Iteration #: 84\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 351s 2ms/step - loss: 0.8724 - acc: 0.7632\n",
      "Generating from seed: p\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q7IEOSbw5EOSqEOS589fqh^zyEOS0AEOSzz$5EOS4EOSEOS05uTsEOSFiEOSEOSp7\n",
      "==================================================\n",
      "Iteration #: 85\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 344s 2ms/step - loss: 0.8684 - acc: 0.7650\n",
      "Generating from seed: 7\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q7IEOSbw5EOSqEOS589fqh^zyEOS0AEOSzz$5EOS4EOSEOS05uTsEOSFiEOSEOSp7Z\n",
      "==================================================\n",
      "Iteration #: 86\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 330s 2ms/step - loss: 0.8625 - acc: 0.7681\n",
      "Generating from seed: Z\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q7IEOSbw5EOSqEOS589fqh^zyEOS0AEOSzz$5EOS4EOSEOS05uTsEOSFiEOSEOSp7Zq\n",
      "==================================================\n",
      "Iteration #: 87\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 368s 2ms/step - loss: 0.8646 - acc: 0.7656\n",
      "Generating from seed: q\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q7IEOSbw5EOSqEOS589fqh^zyEOS0AEOSzz$5EOS4EOSEOS05uTsEOSFiEOSEOSp7Zq!\n",
      "==================================================\n",
      "Iteration #: 88\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 358s 2ms/step - loss: 0.8603 - acc: 0.7681\n",
      "Generating from seed: !\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q7IEOSbw5EOSqEOS589fqh^zyEOS0AEOSzz$5EOS4EOSEOS05uTsEOSFiEOSEOSp7Zq!h\n",
      "==================================================\n",
      "Iteration #: 89\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 393s 2ms/step - loss: 0.8588 - acc: 0.7671\n",
      "Generating from seed: h\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q7IEOSbw5EOSqEOS589fqh^zyEOS0AEOSzz$5EOS4EOSEOS05uTsEOSFiEOSEOSp7Zq!hEOS\n",
      "==================================================\n",
      "Iteration #: 90\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 389s 2ms/step - loss: 0.8533 - acc: 0.7689\n",
      "Generating from seed: EOS\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q7IEOSbw5EOSqEOS589fqh^zyEOS0AEOSzz$5EOS4EOSEOS05uTsEOSFiEOSEOSp7Zq!hEOSS\n",
      "==================================================\n",
      "Iteration #: 91\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 384s 2ms/step - loss: 0.8521 - acc: 0.7698\n",
      "Generating from seed: S\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q7IEOSbw5EOSqEOS589fqh^zyEOS0AEOSzz$5EOS4EOSEOS05uTsEOSFiEOSEOSp7Zq!hEOSSEOS\n",
      "==================================================\n",
      "Iteration #: 92\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 343s 2ms/step - loss: 0.8526 - acc: 0.7695\n",
      "Generating from seed: EOS\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q7IEOSbw5EOSqEOS589fqh^zyEOS0AEOSzz$5EOS4EOSEOS05uTsEOSFiEOSEOSp7Zq!hEOSSEOS1\n",
      "==================================================\n",
      "Iteration #: 93\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 331s 2ms/step - loss: 0.8518 - acc: 0.7690\n",
      "Generating from seed: 1\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q7IEOSbw5EOSqEOS589fqh^zyEOS0AEOSzz$5EOS4EOSEOS05uTsEOSFiEOSEOSp7Zq!hEOSSEOS1y\n",
      "==================================================\n",
      "Iteration #: 94\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 337s 2ms/step - loss: 0.8467 - acc: 0.7707\n",
      "Generating from seed: y\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q7IEOSbw5EOSqEOS589fqh^zyEOS0AEOSzz$5EOS4EOSEOS05uTsEOSFiEOSEOSp7Zq!hEOSSEOS1yEOS\n",
      "==================================================\n",
      "Iteration #: 95\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 324s 2ms/step - loss: 0.8474 - acc: 0.7704\n",
      "Generating from seed: EOS\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q7IEOSbw5EOSqEOS589fqh^zyEOS0AEOSzz$5EOS4EOSEOS05uTsEOSFiEOSEOSp7Zq!hEOSSEOS1yEOS#\n",
      "==================================================\n",
      "Iteration #: 96\n",
      "Epoch 1/1\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "192063/192063 [==============================] - 324s 2ms/step - loss: 0.8487 - acc: 0.7698\n",
      "Generating from seed: #\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q7IEOSbw5EOSqEOS589fqh^zyEOS0AEOSzz$5EOS4EOSEOS05uTsEOSFiEOSEOSp7Zq!hEOSSEOS1yEOS#EOS\n",
      "==================================================\n",
      "Iteration #: 97\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 329s 2ms/step - loss: 0.8470 - acc: 0.7703\n",
      "Generating from seed: EOS\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q7IEOSbw5EOSqEOS589fqh^zyEOS0AEOSzz$5EOS4EOSEOS05uTsEOSFiEOSEOSp7Zq!hEOSSEOS1yEOS#EOSEOS\n",
      "==================================================\n",
      "Iteration #: 98\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 331s 2ms/step - loss: 0.8403 - acc: 0.7730\n",
      "Generating from seed: EOS\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q7IEOSbw5EOSqEOS589fqh^zyEOS0AEOSzz$5EOS4EOSEOS05uTsEOSFiEOSEOSp7Zq!hEOSSEOS1yEOS#EOSEOSEOS\n",
      "==================================================\n",
      "Iteration #: 99\n",
      "Epoch 1/1\n",
      "192063/192063 [==============================] - 330s 2ms/step - loss: 0.8462 - acc: 0.7705\n",
      "Generating from seed: EOS\n",
      "SOS·yyy1sdddEOSf-sdhdfqLsyhEOS755dyfth55555535357Q7IEOSbw5EOSqEOS589fqh^zyEOS0AEOSzz$5EOS4EOSEOS05uTsEOSFiEOSEOSp7Zq!hEOSSEOS1yEOS#EOSEOSEOSEOS\n"
     ]
    }
   ],
   "source": [
    "NUM_ITERATIONS = 10\n",
    "NUM_EPOCHS_PER_ITERATION = 1\n",
    "NUM_PREDS_PER_EPOCH = 100\n",
    "\n",
    "sentence = ''\n",
    "test_idx = np.random.randint(len(chr2idx))\n",
    "test_chars = seed = idx2chr[test_idx]\n",
    "for iteration in range(NUM_ITERATIONS):\n",
    "    print(\"=\" * 50)\n",
    "    print(\"Iteration #: %d\" %(iteration))\n",
    "    model.fit(X, y, batch_size=BATCH_SIZE, epochs=NUM_EPOCHS_PER_ITERATION)\n",
    "    print(\"Generating from seed: %s\" %(test_chars))\n",
    "    for i in range(NUM_EPOCHS_PER_ITERATION):\n",
    "        X_test = np.zeros((1, SEQLEN, len(chr2idx)))\n",
    "        for j, ch in enumerate(test_chars):\n",
    "            X_test[0, j, chr2idx[ch]] = 1\n",
    "        pred = model.predict(X_test, verbose=0)[0]\n",
    "        ypred = idx2chr[np.argmax(pred)]\n",
    "        sentence += ypred\n",
    "        test_chars = ypred\n",
    "    print(seed + sentence)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-25T10:56:43.207617Z",
     "start_time": "2018-11-25T10:56:43.136471Z"
    }
   },
   "outputs": [],
   "source": [
    "PATH_NORMAL = './数据集/ADFA-LD/Training_Data_Master/'\n",
    "PATH_ATTACK = './数据集/ADFA-LD/Attack_Data_Master/Web_Shell_*/'\n",
    "\n",
    "def build_vocab(path1, path2):\n",
    "    fileList1 = glob.glob(path1 + '*.txt')\n",
    "    fileList2 = glob.glob(path2 + '*.txt')\n",
    "    fileList = fileList1 + fileList2\n",
    "    counter = Counter()\n",
    "    for file in fileList:\n",
    "        f = open(file, 'r')\n",
    "        for line in f:\n",
    "            for integer in str(line).strip().split():\n",
    "                counter[integer] += 1\n",
    "    int2idx = {integer:i+2 for i, integer in enumerate(counter)}\n",
    "    int2idx['PAD'] = 0\n",
    "    int2idx['UNK'] = 1\n",
    "    idx2int = {v:k for k, v in int2idx.items()}\n",
    "    return int2idx, idx2int"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-25T10:56:43.241102Z",
     "start_time": "2018-11-25T10:56:43.217013Z"
    }
   },
   "outputs": [],
   "source": [
    "def get_maxlen(path1, path2):\n",
    "    sent_len = 0\n",
    "    sent_maxlen = 0\n",
    "    fileList1 = glob.glob(path1 + '*.txt')\n",
    "    fileList2 = glob.glob(path2 + '*.txt')\n",
    "    fileList = fileList1 + fileList2\n",
    "    for file in fileList:\n",
    "        f = open(file, 'r')\n",
    "        for line in f:\n",
    "            sent_len = len(str(line).strip())\n",
    "            if sent_len > sent_maxlen:\n",
    "                sent_maxlen = sent_len\n",
    "    return sent_maxlen"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-25T10:56:43.478792Z",
     "start_time": "2018-11-25T10:56:43.450879Z"
    }
   },
   "outputs": [],
   "source": [
    "def vectorize(path1, path2, int2idx):\n",
    "    sentences = []\n",
    "    fileList1 = glob.glob(path1 + '*.txt')\n",
    "    fileList2 = glob.glob(path2 + '*.txt')\n",
    "    fileList = fileList1 + fileList2\n",
    "    i = 0\n",
    "    j = 0\n",
    "    for file in fileList:\n",
    "        f = open(file, 'r')\n",
    "        for line in f:\n",
    "            sentence = []\n",
    "            for integer in str(line).strip().split():\n",
    "                sentence.append(int2idx[integer])\n",
    "            sentences.append(sentence)\n",
    "            if file not in fileList2:\n",
    "                i += 1\n",
    "            else:\n",
    "                j += 1\n",
    "    X = pad_sequences(sentences, maxlen=get_maxlen(path1, path2))\n",
    "    Y = to_categorical(i * [0] + j * [1], num_classes=2)\n",
    "    return X, Y  "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-25T11:01:33.630830Z",
     "start_time": "2018-11-25T11:01:29.329232Z"
    }
   },
   "outputs": [],
   "source": [
    "int2idx, idx2int = build_vocab(PATH_NORMAL, PATH_ATTACK)\n",
    "maxlen = get_maxlen(PATH_NORMAL, PATH_ATTACK)\n",
    "X, Y = vectorize(PATH_NORMAL, PATH_ATTACK, int2idx)\n",
    "X_train, X_test, Y_train, Y_test = train_test_split(X, Y, test_size=0.2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-25T11:03:56.503523Z",
     "start_time": "2018-11-25T11:03:55.897551Z"
    }
   },
   "outputs": [],
   "source": [
    "EMBEDDING_SIZE = 128\n",
    "HIDDEN_LAYER_SIZE = 64\n",
    "BATCH_SIZE = 256\n",
    "NUM_EPOCHS = 10\n",
    "\n",
    "model = Sequential()\n",
    "model.add(Embedding(len(int2idx), EMBEDDING_SIZE, input_length=maxlen))\n",
    "model.add(SpatialDropout1D(0.2))\n",
    "model.add(LSTM(HIDDEN_LAYER_SIZE, dropout=0.2, recurrent_dropout=0.2))\n",
    "model.add(Dense(2, activation=\"sigmoid\"))\n",
    "\n",
    "model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "start_time": "2018-11-25T11:03:57.858Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 760 samples, validate on 191 samples\n",
      "Epoch 1/10\n",
      "256/760 [=========>....................] - ETA: 12:41 - loss: 0.6976 - acc: 0.3398"
     ]
    }
   ],
   "source": [
    "history = model.fit(X_train, Y_train, batch_size=BATCH_SIZE, epochs=NUM_EPOCHS, validation_data=(X_test, Y_test))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "start_time": "2018-11-25T11:04:33.842Z"
    }
   },
   "outputs": [],
   "source": [
    "model.evaluate(X_test, Y_test, batch_size=BATCH_SIZE)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-25T11:30:15.076739Z",
     "start_time": "2018-11-25T11:30:15.044953Z"
    }
   },
   "outputs": [],
   "source": [
    "def build_vocab(path):\n",
    "    fileList = glob.glob(path + 'User*')\n",
    "    counter = Counter()\n",
    "    for file in fileList:\n",
    "        user = []\n",
    "        f = open(file, 'r')\n",
    "        for line in f:\n",
    "            for word in line.strip().split():\n",
    "                counter[word.lower()] += 1\n",
    "    word2idx = {w:i+2 for i, w in enumerate(counter.most_common())}\n",
    "    word2idx['PAD'] = 0\n",
    "    word2idx['UNK'] = 1\n",
    "    idx2word = {v:k for k, v in word2idx.items()}\n",
    "    return word2idx, idx2word"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-25T11:30:15.094313Z",
     "start_time": "2018-11-25T11:30:15.084748Z"
    }
   },
   "outputs": [],
   "source": [
    "def get_maxlen(path):\n",
    "    fileList = glob.glob(path + 'User*')\n",
    "    command_maxlen = 0\n",
    "    for file in fileList:\n",
    "        user = []\n",
    "        f = open(file, 'r')\n",
    "        for line in f:\n",
    "            user.append(line.strip())\n",
    "        command_len = len(user)\n",
    "        if command_len > command_maxlen:\n",
    "            command_maxlen = command_len\n",
    "    return command_maxlen"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-25T11:30:15.107980Z",
     "start_time": "2018-11-25T11:30:15.097528Z"
    }
   },
   "outputs": [],
   "source": [
    "def vectorize(path, word2idx):\n",
    "    users = []\n",
    "    fileList = glob.glob(path + 'User*')\n",
    "    for file in fileList:\n",
    "        user = []\n",
    "        f = open(file, 'r')\n",
    "        for line in f:\n",
    "            try:\n",
    "                user.append(word2idx[line.strip()])\n",
    "            except KeyError:\n",
    "                user.append(word2idx['UNK'])                \n",
    "        users.append(user)\n",
    "    return pad_sequences(users, maxlen=get_maxlen(path))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-25T11:30:15.122557Z",
     "start_time": "2018-11-25T11:30:15.112019Z"
    }
   },
   "outputs": [],
   "source": [
    "def get_label(filename):\n",
    "    label = np.loadtxt(filename)\n",
    "    zero = np.zeros((50, label.shape[1]))\n",
    "    labels = np.concatenate([zero, label])\n",
    "    mean = labels.mean()\n",
    "    y = []\n",
    "    for i in np.sum(labels, axis=0):\n",
    "        if i > mean:\n",
    "            y.append(1)\n",
    "        else:\n",
    "            y.append(0)\n",
    "    return y"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-25T11:30:18.673759Z",
     "start_time": "2018-11-25T11:30:15.127208Z"
    }
   },
   "outputs": [],
   "source": [
    "PATH = './数据集/masquerade/'\n",
    "FILENAME = './数据集/masquerade/labels.txt'\n",
    "word2idx, idx2word = build_vocab(PATH)\n",
    "maxlen = get_maxlen(PATH)\n",
    "X = vectorize(PATH, word2idx)\n",
    "Y = to_categorical(get_label(FILENAME), num_classes=2)\n",
    "X_train, X_test, Y_train, Y_test = train_test_split(X, Y, test_size=0.2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-11-25T11:30:19.878297Z",
     "start_time": "2018-11-25T11:30:18.682572Z"
    }
   },
   "outputs": [],
   "source": [
    "EMBEDDING_SIZE = 128\n",
    "HIDDEN_LAYER_SIZE = 64\n",
    "BATCH_SIZE = 256\n",
    "NUM_EPOCHS = 10\n",
    "\n",
    "model = Sequential()\n",
    "model.add(Embedding(len(word2idx), EMBEDDING_SIZE, input_length=maxlen))\n",
    "model.add(SpatialDropout1D(0.2))\n",
    "model.add(LSTM(HIDDEN_LAYER_SIZE, dropout=0.2, recurrent_dropout=0.2))\n",
    "model.add(Dense(2, activation=\"sigmoid\"))\n",
    "\n",
    "model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "start_time": "2018-11-25T11:30:15.162Z"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 40 samples, validate on 10 samples\n",
      "Epoch 1/10\n",
      "40/40 [==============================] - 128s 3s/step - loss: 0.6916 - acc: 0.5250 - val_loss: 0.6980 - val_acc: 0.4000\n",
      "Epoch 2/10\n",
      "40/40 [==============================] - 89s 2s/step - loss: 0.6879 - acc: 0.6250 - val_loss: 0.7018 - val_acc: 0.4000\n",
      "Epoch 3/10\n",
      "40/40 [==============================] - 79s 2s/step - loss: 0.6844 - acc: 0.6250 - val_loss: 0.7056 - val_acc: 0.4000\n",
      "Epoch 4/10\n",
      "40/40 [==============================] - 86s 2s/step - loss: 0.6781 - acc: 0.6250 - val_loss: 0.7099 - val_acc: 0.4000\n",
      "Epoch 5/10\n",
      "40/40 [==============================] - 87s 2s/step - loss: 0.6777 - acc: 0.6250 - val_loss: 0.7150 - val_acc: 0.4000\n",
      "Epoch 6/10\n",
      "40/40 [==============================] - 86s 2s/step - loss: 0.6745 - acc: 0.6250 - val_loss: 0.7206 - val_acc: 0.4000\n",
      "Epoch 7/10\n",
      "40/40 [==============================] - 113s 3s/step - loss: 0.6692 - acc: 0.6250 - val_loss: 0.7274 - val_acc: 0.4000\n",
      "Epoch 8/10\n",
      "40/40 [==============================] - 95s 2s/step - loss: 0.6721 - acc: 0.6250 - val_loss: 0.7350 - val_acc: 0.4000\n",
      "Epoch 9/10\n"
     ]
    }
   ],
   "source": [
    "history = model.fit(X_train, Y_train, batch_size=BATCH_SIZE, epochs=NUM_EPOCHS, validation_data=(X_test, Y_test))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "ExecuteTime": {
     "start_time": "2018-11-25T11:30:16.969Z"
    }
   },
   "outputs": [],
   "source": [
    "model.evaluate(X_test, Y_test, batch_size=BATCH_SIZE)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}