mirrornerror/hyperopt_gen_simple.ipynb

## hyperopt_gen_simple.ipynb
{
  "cells": [
    {
      "metadata": {},
      "cell_type": "markdown",
      "source": "## Kaggel: Digit Recognizer(MNIST)  by Hyperopt + fit_generator()\nKaggle Digit recognizer: https://www.kaggle.com/c/digit-recognizer  \nHyperopt: https://github.com/hyperopt/hyperopt  \n\n### Best Score: 0.99771\n* max_evals= 10 (time: 1h 33m), score: 0.99771\n* max_evals= 20 (time: --),  score: \n* max_evals= 30 (time: --),  score: \n\nPython 3.6  \nNVIDIA GTX1060  "
    },
    {
      "metadata": {
        "ExecuteTime": {
          "start_time": "2018-11-20T11:14:22.856808Z",
          "end_time": "2018-11-20T11:14:22.896224Z"
        },
        "trusted": true
      },
      "cell_type": "code",
      "source": "import warnings\nwarnings.filterwarnings('ignore')\n\nimport hyperopt\nfrom hyperopt import hp, fmin, rand, tpe, Trials, space_eval, STATUS_OK\n\nfrom keras.models import Sequential\nfrom keras.layers import Dense, Dropout, Flatten, Conv2D, MaxPool2D, BatchNormalization\nfrom keras.callbacks import EarlyStopping, ModelCheckpoint, ReduceLROnPlateau\nfrom keras.preprocessing.image import ImageDataGenerator\nimport keras\n\nimport pandas as pd\nfrom sklearn.model_selection import train_test_split\nimport numpy as np\nimport matplotlib.pyplot as plt\n%matplotlib inline\n\n# fix random seed\nimport tensorflow as tf\nimport random as rn\nimport os\nos.environ['PYTHONHASHSEED'] = '0'\nseed = 123\nrn.seed(seed)\nnp.random.seed(seed)\nsession_conf = tf.ConfigProto(intra_op_parallelism_threads=1, inter_op_parallelism_threads=1)\nfrom keras import backend as K\ntf.set_random_seed(seed)\nsess = tf.Session(graph=tf.get_default_graph(), config=session_conf)\nK.set_session(sess)\n\nprint('tensorflow ver.:',tf.__version__)\nprint('keras ver.     : ', keras.__version__)\nprint('hyperopt ver.  :   ', hyperopt.__version__)",
      "execution_count": 97,
      "outputs": [
        {
          "output_type": "stream",
          "text": "tensorflow ver.: 1.11.0\nkeras ver.     :  2.2.2\nhyperopt ver.  :    0.2\n",
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {},
      "cell_type": "markdown",
      "source": "## Data preparation: MNIST from Kaggle\n* data split proportion: test_size=0.15"
    },
    {
      "metadata": {
        "ExecuteTime": {
          "start_time": "2018-11-20T11:14:22.897852Z",
          "end_time": "2018-11-20T11:14:25.747671Z"
        },
        "trusted": true
      },
      "cell_type": "code",
      "source": "train = pd.read_csv('../train.csv')\nlabel = train.label\ntrain = train.drop(['label'], axis=1)\ntrain = train.values.reshape(-1, 28, 28, 1)\n\nx_train, x_test, y_train, y_test = train_test_split(train, label, test_size=0.15, shuffle=True, random_state=seed)\nx_train, x_val, y_train, y_val = train_test_split(train, label, test_size=0.15, shuffle=True, random_state=seed)\nx_train = x_train.astype('float32') / 255.0\nx_test = x_test.astype('float32') / 255.0\nx_val = x_val.astype('float32') / 255.0\n\n[x_train.shape, x_val.shape, x_test.shape]",
      "execution_count": 98,
      "outputs": [
        {
          "output_type": "execute_result",
          "execution_count": 98,
          "data": {
            "text/plain": "[(35700, 28, 28, 1), (6300, 28, 28, 1), (6300, 28, 28, 1)]"
          },
          "metadata": {}
        }
      ]
    },
    {
      "metadata": {},
      "cell_type": "markdown",
      "source": "## Hyperparameters for Hyperopt:  \n*  Dropout rates"
    },
    {
      "metadata": {
        "ExecuteTime": {
          "start_time": "2018-11-20T11:14:25.749262Z",
          "end_time": "2018-11-20T11:14:25.752837Z"
        },
        "trusted": true
      },
      "cell_type": "code",
      "source": "params = {\n    'Dropout_0': hp.uniform('Dropout_0', 0.0, 0.5),\n    'Dropout_1': hp.uniform('Dropout_1', 0.0, 0.5),\n    'Dropout_2': hp.uniform('Dropout_2', 0.0, 0.5)\n}",
      "execution_count": 99,
      "outputs": []
    },
    {
      "metadata": {},
      "cell_type": "markdown",
      "source": "## CNN Model:\n* return a dictionary includes at least 'loss' and 'status' for hyperopt\n* the model pass -acc(negative accuracy) as a 'loss' value to the fmin()\n* data augmentation: to use fit_generator() instead of fit()"
    },
    {
      "metadata": {
        "ExecuteTime": {
          "start_time": "2018-11-20T11:14:25.754483Z",
          "end_time": "2018-11-20T11:14:25.841489Z"
        },
        "trusted": true
      },
      "cell_type": "code",
      "source": "batch_size = 32\n\ncnt = 0\ndef cnn_model(params):\n    \n    initializer = keras.initializers.glorot_uniform(seed=seed)\n    \n    model = Sequential()  \n    model.add(Conv2D(32, 3, activation='relu', kernel_initializer=initializer, input_shape=(28,28,1)))\n    model.add(BatchNormalization())\n    model.add(Conv2D(32, 3, activation='relu', kernel_initializer=initializer))\n    model.add(BatchNormalization())\n    model.add(Conv2D(32, 4, strides=2, padding='same', activation='relu', kernel_initializer=initializer))\n    model.add(BatchNormalization())\n    model.add(Dropout(params['Dropout_0'], seed=seed))\n    \n    model.add(Conv2D(64, 3, activation='relu', kernel_initializer=initializer))\n    model.add(BatchNormalization())\n    model.add(Conv2D(64, 3, activation='relu', kernel_initializer=initializer))\n    model.add(BatchNormalization())\n    model.add(Conv2D(64, 4, strides=2, padding='same', activation='relu', kernel_initializer=initializer))\n    model.add(BatchNormalization())\n    model.add(Dropout(params['Dropout_1'], seed=seed))\n\n    model.add(Flatten())\n    model.add(Dense(256, activation=\"relu\", kernel_initializer=initializer))\n    model.add(BatchNormalization())\n    model.add(Dropout(params['Dropout_2'], seed=seed))\n    \n    model.add(Dense(10, activation = \"softmax\", kernel_initializer=initializer))\n\n    model.compile(loss='sparse_categorical_crossentropy', optimizer='adam', metrics=['accuracy'])\n    \n    datagen = ImageDataGenerator(rotation_range=10,\n                                 width_shift_range=0.1,\n                                 height_shift_range=0.1,\n                                 zoom_range=0.1)\n\n    datagen.fit(x_train, seed=seed)\n\n    reduce_lr = ReduceLROnPlateau(monitor='val_loss', factor=0.5, patience=2, min_lr=1e-6, verbose=1)\n    early_stop = EarlyStopping(monitor='val_loss', patience=5, verbose=1)\n\n    hist = model.fit_generator(datagen.flow(x_train, y_train, batch_size=batch_size), \n                               epochs=50,\n                               validation_data=(x_val, y_val),\n                               steps_per_epoch=x_train.shape[0]//batch_size,\n                               callbacks=[reduce_lr, early_stop],\n                               verbose=2)\n\n    loss = hist.history['val_loss'][-1]\n    acc = hist.history['val_acc'][-1]\n\n    global cnt\n    print(cnt, ': Val loss:', loss, ': Val acc:', acc, '\\n\\n')\n    cnt += 1\n    \n    return {'loss': -acc, 'status': STATUS_OK, 'model': model, 'hist': hist}\n    ",
      "execution_count": 100,
      "outputs": []
    },
    {
      "metadata": {},
      "cell_type": "markdown",
      "source": "## Search the best Hyperparameters & the best model:"
    },
    {
      "metadata": {
        "ExecuteTime": {
          "start_time": "2018-11-20T11:14:25.842937Z",
          "end_time": "2018-11-20T12:47:30.147188Z"
        },
        "scrolled": true,
        "trusted": true
      },
      "cell_type": "code",
      "source": "trials = Trials()\nbest = fmin(fn=cnn_model, \n            space=params, \n            algo=tpe.suggest, \n            max_evals=10, # 50: 5h 15m\n            trials=trials,\n            verbose=1,\n            rstate=np.random.RandomState(seed))\n\nbest",
      "execution_count": 101,
      "outputs": [
        {
          "output_type": "stream",
          "text": "Epoch 1/50\n - 27s - loss: 0.3667 - acc: 0.8860 - val_loss: 0.0668 - val_acc: 0.9810\nEpoch 2/50\n - 18s - loss: 0.1195 - acc: 0.9632 - val_loss: 0.0747 - val_acc: 0.9784\nEpoch 3/50\n - 18s - loss: 0.0975 - acc: 0.9699 - val_loss: 0.0300 - val_acc: 0.9905\nEpoch 4/50\n - 18s - loss: 0.0805 - acc: 0.9753 - val_loss: 0.0336 - val_acc: 0.9883\nEpoch 5/50\n - 18s - loss: 0.0763 - acc: 0.9767 - val_loss: 0.0271 - val_acc: 0.9913\nEpoch 6/50\n - 18s - loss: 0.0695 - acc: 0.9788 - val_loss: 0.0300 - val_acc: 0.9902\nEpoch 7/50\n - 19s - loss: 0.0629 - acc: 0.9821 - val_loss: 0.0250 - val_acc: 0.9919\nEpoch 8/50\n - 18s - loss: 0.0586 - acc: 0.9820 - val_loss: 0.0196 - val_acc: 0.9937\nEpoch 9/50\n - 18s - loss: 0.0525 - acc: 0.9843 - val_loss: 0.0194 - val_acc: 0.9925\nEpoch 10/50\n - 17s - loss: 0.0510 - acc: 0.9849 - val_loss: 0.0216 - val_acc: 0.9929\nEpoch 11/50\n - 18s - loss: 0.0480 - acc: 0.9858 - val_loss: 0.0199 - val_acc: 0.9925\n\nEpoch 00011: ReduceLROnPlateau reducing learning rate to 0.0005000000237487257.\nEpoch 12/50\n - 18s - loss: 0.0345 - acc: 0.9895 - val_loss: 0.0145 - val_acc: 0.9949\nEpoch 13/50\n - 18s - loss: 0.0327 - acc: 0.9903 - val_loss: 0.0132 - val_acc: 0.9959\nEpoch 14/50\n - 17s - loss: 0.0323 - acc: 0.9909 - val_loss: 0.0146 - val_acc: 0.9956\nEpoch 15/50\n - 18s - loss: 0.0320 - acc: 0.9904 - val_loss: 0.0121 - val_acc: 0.9963\nEpoch 16/50\n - 17s - loss: 0.0281 - acc: 0.9913 - val_loss: 0.0098 - val_acc: 0.9968\nEpoch 17/50\n - 18s - loss: 0.0275 - acc: 0.9917 - val_loss: 0.0131 - val_acc: 0.9967\nEpoch 18/50\n - 18s - loss: 0.0255 - acc: 0.9921 - val_loss: 0.0110 - val_acc: 0.9970\n\nEpoch 00018: ReduceLROnPlateau reducing learning rate to 0.0002500000118743628.\nEpoch 19/50\n - 17s - loss: 0.0225 - acc: 0.9928 - val_loss: 0.0096 - val_acc: 0.9965\nEpoch 20/50\n - 17s - loss: 0.0202 - acc: 0.9939 - val_loss: 0.0112 - val_acc: 0.9963\nEpoch 21/50\n - 17s - loss: 0.0208 - acc: 0.9941 - val_loss: 0.0123 - val_acc: 0.9967\n\nEpoch 00021: ReduceLROnPlateau reducing learning rate to 0.0001250000059371814.\nEpoch 22/50\n - 17s - loss: 0.0185 - acc: 0.9946 - val_loss: 0.0101 - val_acc: 0.9975\nEpoch 23/50\n - 18s - loss: 0.0163 - acc: 0.9948 - val_loss: 0.0102 - val_acc: 0.9975\n\nEpoch 00023: ReduceLROnPlateau reducing learning rate to 6.25000029685907e-05.\nEpoch 24/50\n - 17s - loss: 0.0141 - acc: 0.9954 - val_loss: 0.0105 - val_acc: 0.9975\nEpoch 00024: early stopping\n0 : Val loss: 0.010491497190595264 : Val acc: 0.9974603174603175 \n\n\nEpoch 1/50\n - 27s - loss: 0.3405 - acc: 0.8966 - val_loss: 0.0632 - val_acc: 0.9813\nEpoch 2/50\n - 18s - loss: 0.1191 - acc: 0.9633 - val_loss: 0.0848 - val_acc: 0.9746\nEpoch 3/50\n - 18s - loss: 0.0927 - acc: 0.9721 - val_loss: 0.0311 - val_acc: 0.9908\nEpoch 4/50\n - 18s - loss: 0.0780 - acc: 0.9755 - val_loss: 0.0376 - val_acc: 0.9886\nEpoch 5/50\n - 18s - loss: 0.0743 - acc: 0.9770 - val_loss: 0.0528 - val_acc: 0.9844\n\nEpoch 00005: ReduceLROnPlateau reducing learning rate to 0.0005000000237487257.\nEpoch 6/50\n - 18s - loss: 0.0493 - acc: 0.9856 - val_loss: 0.0268 - val_acc: 0.9916\nEpoch 7/50\n - 18s - loss: 0.0450 - acc: 0.9872 - val_loss: 0.0253 - val_acc: 0.9913\nEpoch 8/50\n - 18s - loss: 0.0436 - acc: 0.9868 - val_loss: 0.0255 - val_acc: 0.9917\nEpoch 9/50\n - 18s - loss: 0.0411 - acc: 0.9881 - val_loss: 0.0180 - val_acc: 0.9940\nEpoch 10/50\n - 18s - loss: 0.0406 - acc: 0.9880 - val_loss: 0.0174 - val_acc: 0.9944\nEpoch 11/50\n - 18s - loss: 0.0389 - acc: 0.9886 - val_loss: 0.0161 - val_acc: 0.9954\nEpoch 12/50\n - 18s - loss: 0.0336 - acc: 0.9899 - val_loss: 0.0165 - val_acc: 0.9946\nEpoch 13/50\n - 18s - loss: 0.0335 - acc: 0.9900 - val_loss: 0.0158 - val_acc: 0.9943\nEpoch 14/50\n - 18s - loss: 0.0339 - acc: 0.9901 - val_loss: 0.0148 - val_acc: 0.9948\nEpoch 15/50\n - 18s - loss: 0.0340 - acc: 0.9899 - val_loss: 0.0141 - val_acc: 0.9954\nEpoch 16/50\n - 18s - loss: 0.0300 - acc: 0.9909 - val_loss: 0.0194 - val_acc: 0.9951\nEpoch 17/50\n - 18s - loss: 0.0302 - acc: 0.9908 - val_loss: 0.0126 - val_acc: 0.9968\nEpoch 18/50\n - 18s - loss: 0.0290 - acc: 0.9915 - val_loss: 0.0123 - val_acc: 0.9956\nEpoch 19/50\n - 18s - loss: 0.0302 - acc: 0.9911 - val_loss: 0.0149 - val_acc: 0.9962\nEpoch 20/50\n - 18s - loss: 0.0266 - acc: 0.9917 - val_loss: 0.0129 - val_acc: 0.9957\n\nEpoch 00020: ReduceLROnPlateau reducing learning rate to 0.0002500000118743628.\nEpoch 21/50\n - 18s - loss: 0.0212 - acc: 0.9936 - val_loss: 0.0119 - val_acc: 0.9963\nEpoch 22/50\n - 18s - loss: 0.0196 - acc: 0.9945 - val_loss: 0.0120 - val_acc: 0.9963\nEpoch 23/50\n - 18s - loss: 0.0185 - acc: 0.9944 - val_loss: 0.0108 - val_acc: 0.9968\nEpoch 24/50\n - 18s - loss: 0.0164 - acc: 0.9952 - val_loss: 0.0151 - val_acc: 0.9948\nEpoch 25/50\n - 18s - loss: 0.0187 - acc: 0.9945 - val_loss: 0.0175 - val_acc: 0.9954\n\nEpoch 00025: ReduceLROnPlateau reducing learning rate to 0.0001250000059371814.\nEpoch 26/50\n - 18s - loss: 0.0161 - acc: 0.9951 - val_loss: 0.0117 - val_acc: 0.9965\nEpoch 27/50\n - 18s - loss: 0.0140 - acc: 0.9956 - val_loss: 0.0117 - val_acc: 0.9962\n\nEpoch 00027: ReduceLROnPlateau reducing learning rate to 6.25000029685907e-05.\nEpoch 28/50\n - 18s - loss: 0.0134 - acc: 0.9963 - val_loss: 0.0123 - val_acc: 0.9965\nEpoch 00028: early stopping\n1 : Val loss: 0.012333203735673572 : Val acc: 0.9965079365079365 \n\n\nEpoch 1/50\n - 28s - loss: 0.5129 - acc: 0.8437 - val_loss: 0.0584 - val_acc: 0.9833\nEpoch 2/50\n - 18s - loss: 0.1607 - acc: 0.9511 - val_loss: 0.0569 - val_acc: 0.9824\nEpoch 3/50\n - 18s - loss: 0.1268 - acc: 0.9620 - val_loss: 0.0320 - val_acc: 0.9892\nEpoch 4/50\n - 18s - loss: 0.1002 - acc: 0.9696 - val_loss: 0.0408 - val_acc: 0.9871\nEpoch 5/50\n - 18s - loss: 0.0941 - acc: 0.9723 - val_loss: 0.0357 - val_acc: 0.9895\n\nEpoch 00005: ReduceLROnPlateau reducing learning rate to 0.0005000000237487257.\nEpoch 6/50\n - 18s - loss: 0.0727 - acc: 0.9788 - val_loss: 0.0237 - val_acc: 0.9929\nEpoch 7/50\n - 18s - loss: 0.0615 - acc: 0.9823 - val_loss: 0.0227 - val_acc: 0.9938\nEpoch 8/50\n - 18s - loss: 0.0574 - acc: 0.9823 - val_loss: 0.0201 - val_acc: 0.9933\nEpoch 9/50\n - 18s - loss: 0.0565 - acc: 0.9823 - val_loss: 0.0182 - val_acc: 0.9944\nEpoch 10/50\n - 18s - loss: 0.0565 - acc: 0.9833 - val_loss: 0.0223 - val_acc: 0.9933\nEpoch 11/50\n - 18s - loss: 0.0520 - acc: 0.9846 - val_loss: 0.0166 - val_acc: 0.9938\nEpoch 12/50\n - 18s - loss: 0.0497 - acc: 0.9850 - val_loss: 0.0184 - val_acc: 0.9943\nEpoch 13/50\n - 18s - loss: 0.0490 - acc: 0.9856 - val_loss: 0.0152 - val_acc: 0.9957\nEpoch 14/50\n - 18s - loss: 0.0464 - acc: 0.9863 - val_loss: 0.0187 - val_acc: 0.9933\nEpoch 15/50\n - 18s - loss: 0.0450 - acc: 0.9871 - val_loss: 0.0141 - val_acc: 0.9954\nEpoch 16/50\n - 19s - loss: 0.0446 - acc: 0.9860 - val_loss: 0.0142 - val_acc: 0.9952\nEpoch 17/50\n - 18s - loss: 0.0416 - acc: 0.9877 - val_loss: 0.0167 - val_acc: 0.9932\n\nEpoch 00017: ReduceLROnPlateau reducing learning rate to 0.0002500000118743628.\nEpoch 18/50\n - 19s - loss: 0.0373 - acc: 0.9891 - val_loss: 0.0114 - val_acc: 0.9967\nEpoch 19/50\n - 18s - loss: 0.0346 - acc: 0.9894 - val_loss: 0.0109 - val_acc: 0.9971\nEpoch 20/50\n - 18s - loss: 0.0327 - acc: 0.9907 - val_loss: 0.0114 - val_acc: 0.9970\nEpoch 21/50\n - 18s - loss: 0.0295 - acc: 0.9915 - val_loss: 0.0119 - val_acc: 0.9968\n\nEpoch 00021: ReduceLROnPlateau reducing learning rate to 0.0001250000059371814.\nEpoch 22/50\n - 18s - loss: 0.0285 - acc: 0.9914 - val_loss: 0.0102 - val_acc: 0.9975\nEpoch 23/50\n - 18s - loss: 0.0255 - acc: 0.9923 - val_loss: 0.0088 - val_acc: 0.9978\nEpoch 24/50\n - 18s - loss: 0.0268 - acc: 0.9918 - val_loss: 0.0097 - val_acc: 0.9973\nEpoch 25/50\n - 18s - loss: 0.0265 - acc: 0.9920 - val_loss: 0.0102 - val_acc: 0.9971\n\nEpoch 00025: ReduceLROnPlateau reducing learning rate to 6.25000029685907e-05.\nEpoch 26/50\n - 18s - loss: 0.0235 - acc: 0.9931 - val_loss: 0.0102 - val_acc: 0.9970\nEpoch 27/50\n - 18s - loss: 0.0230 - acc: 0.9931 - val_loss: 0.0095 - val_acc: 0.9973\n\nEpoch 00027: ReduceLROnPlateau reducing learning rate to 3.125000148429535e-05.\nEpoch 28/50\n - 18s - loss: 0.0228 - acc: 0.9932 - val_loss: 0.0094 - val_acc: 0.9973\nEpoch 00028: early stopping\n2 : Val loss: 0.009427438194656537 : Val acc: 0.9973015873015874 \n\n\nEpoch 1/50\n - 28s - loss: 0.3601 - acc: 0.8890 - val_loss: 0.0557 - val_acc: 0.9833\n",
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": "Epoch 2/50\n - 18s - loss: 0.1263 - acc: 0.9612 - val_loss: 0.0759 - val_acc: 0.9779\nEpoch 3/50\n - 19s - loss: 0.0970 - acc: 0.9703 - val_loss: 0.0315 - val_acc: 0.9897\nEpoch 4/50\n - 19s - loss: 0.0823 - acc: 0.9754 - val_loss: 0.0324 - val_acc: 0.9905\nEpoch 5/50\n - 19s - loss: 0.0780 - acc: 0.9761 - val_loss: 0.0297 - val_acc: 0.9906\nEpoch 6/50\n - 18s - loss: 0.0689 - acc: 0.9797 - val_loss: 0.0284 - val_acc: 0.9905\nEpoch 7/50\n - 18s - loss: 0.0627 - acc: 0.9810 - val_loss: 0.0275 - val_acc: 0.9906\nEpoch 8/50\n - 18s - loss: 0.0557 - acc: 0.9831 - val_loss: 0.0279 - val_acc: 0.9916\nEpoch 9/50\n - 18s - loss: 0.0546 - acc: 0.9844 - val_loss: 0.0189 - val_acc: 0.9943\nEpoch 10/50\n - 18s - loss: 0.0519 - acc: 0.9846 - val_loss: 0.0186 - val_acc: 0.9941\nEpoch 11/50\n - 18s - loss: 0.0490 - acc: 0.9854 - val_loss: 0.0169 - val_acc: 0.9944\nEpoch 12/50\n - 18s - loss: 0.0461 - acc: 0.9864 - val_loss: 0.0226 - val_acc: 0.9937\nEpoch 13/50\n - 18s - loss: 0.0410 - acc: 0.9876 - val_loss: 0.0161 - val_acc: 0.9946\nEpoch 14/50\n - 18s - loss: 0.0416 - acc: 0.9875 - val_loss: 0.0199 - val_acc: 0.9933\nEpoch 15/50\n - 18s - loss: 0.0411 - acc: 0.9879 - val_loss: 0.0147 - val_acc: 0.9954\nEpoch 16/50\n - 18s - loss: 0.0377 - acc: 0.9889 - val_loss: 0.0159 - val_acc: 0.9949\nEpoch 17/50\n - 18s - loss: 0.0363 - acc: 0.9891 - val_loss: 0.0231 - val_acc: 0.9921\n\nEpoch 00017: ReduceLROnPlateau reducing learning rate to 0.0005000000237487257.\nEpoch 18/50\n - 18s - loss: 0.0274 - acc: 0.9918 - val_loss: 0.0108 - val_acc: 0.9968\nEpoch 19/50\n - 19s - loss: 0.0253 - acc: 0.9924 - val_loss: 0.0108 - val_acc: 0.9967\nEpoch 20/50\n - 18s - loss: 0.0239 - acc: 0.9926 - val_loss: 0.0135 - val_acc: 0.9963\n\nEpoch 00020: ReduceLROnPlateau reducing learning rate to 0.0002500000118743628.\nEpoch 21/50\n - 18s - loss: 0.0196 - acc: 0.9943 - val_loss: 0.0122 - val_acc: 0.9962\nEpoch 22/50\n - 18s - loss: 0.0187 - acc: 0.9942 - val_loss: 0.0111 - val_acc: 0.9965\n\nEpoch 00022: ReduceLROnPlateau reducing learning rate to 0.0001250000059371814.\nEpoch 23/50\n - 19s - loss: 0.0160 - acc: 0.9949 - val_loss: 0.0101 - val_acc: 0.9970\nEpoch 24/50\n - 19s - loss: 0.0153 - acc: 0.9953 - val_loss: 0.0097 - val_acc: 0.9970\nEpoch 25/50\n - 18s - loss: 0.0164 - acc: 0.9954 - val_loss: 0.0089 - val_acc: 0.9978\nEpoch 26/50\n - 18s - loss: 0.0151 - acc: 0.9950 - val_loss: 0.0096 - val_acc: 0.9970\nEpoch 27/50\n - 18s - loss: 0.0162 - acc: 0.9952 - val_loss: 0.0087 - val_acc: 0.9968\nEpoch 28/50\n - 18s - loss: 0.0152 - acc: 0.9956 - val_loss: 0.0094 - val_acc: 0.9975\nEpoch 29/50\n - 18s - loss: 0.0149 - acc: 0.9957 - val_loss: 0.0092 - val_acc: 0.9971\n\nEpoch 00029: ReduceLROnPlateau reducing learning rate to 6.25000029685907e-05.\nEpoch 30/50\n - 18s - loss: 0.0132 - acc: 0.9958 - val_loss: 0.0089 - val_acc: 0.9973\nEpoch 31/50\n - 18s - loss: 0.0149 - acc: 0.9956 - val_loss: 0.0080 - val_acc: 0.9975\nEpoch 32/50\n - 18s - loss: 0.0128 - acc: 0.9956 - val_loss: 0.0083 - val_acc: 0.9968\nEpoch 33/50\n - 18s - loss: 0.0147 - acc: 0.9958 - val_loss: 0.0087 - val_acc: 0.9971\n\nEpoch 00033: ReduceLROnPlateau reducing learning rate to 3.125000148429535e-05.\nEpoch 34/50\n - 18s - loss: 0.0131 - acc: 0.9961 - val_loss: 0.0084 - val_acc: 0.9973\nEpoch 35/50\n - 18s - loss: 0.0131 - acc: 0.9959 - val_loss: 0.0086 - val_acc: 0.9971\n\nEpoch 00035: ReduceLROnPlateau reducing learning rate to 1.5625000742147677e-05.\nEpoch 36/50\n - 18s - loss: 0.0124 - acc: 0.9962 - val_loss: 0.0084 - val_acc: 0.9970\nEpoch 00036: early stopping\n3 : Val loss: 0.008394128472741975 : Val acc: 0.996984126984127 \n\n\nEpoch 1/50\n - 28s - loss: 0.3337 - acc: 0.8957 - val_loss: 0.0655 - val_acc: 0.9794\nEpoch 2/50\n - 19s - loss: 0.1130 - acc: 0.9649 - val_loss: 0.0889 - val_acc: 0.9752\nEpoch 3/50\n - 19s - loss: 0.0942 - acc: 0.9713 - val_loss: 0.0281 - val_acc: 0.9919\nEpoch 4/50\n - 19s - loss: 0.0764 - acc: 0.9761 - val_loss: 0.0331 - val_acc: 0.9906\nEpoch 5/50\n - 19s - loss: 0.0727 - acc: 0.9779 - val_loss: 0.0359 - val_acc: 0.9905\n\nEpoch 00005: ReduceLROnPlateau reducing learning rate to 0.0005000000237487257.\nEpoch 6/50\n - 19s - loss: 0.0490 - acc: 0.9855 - val_loss: 0.0251 - val_acc: 0.9927\nEpoch 7/50\n - 19s - loss: 0.0462 - acc: 0.9860 - val_loss: 0.0218 - val_acc: 0.9925\nEpoch 8/50\n - 19s - loss: 0.0411 - acc: 0.9871 - val_loss: 0.0177 - val_acc: 0.9943\nEpoch 9/50\n - 19s - loss: 0.0403 - acc: 0.9878 - val_loss: 0.0139 - val_acc: 0.9957\nEpoch 10/50\n - 19s - loss: 0.0417 - acc: 0.9877 - val_loss: 0.0137 - val_acc: 0.9952\nEpoch 11/50\n - 19s - loss: 0.0361 - acc: 0.9893 - val_loss: 0.0160 - val_acc: 0.9957\nEpoch 12/50\n - 19s - loss: 0.0351 - acc: 0.9889 - val_loss: 0.0146 - val_acc: 0.9949\n\nEpoch 00012: ReduceLROnPlateau reducing learning rate to 0.0002500000118743628.\nEpoch 13/50\n - 19s - loss: 0.0282 - acc: 0.9909 - val_loss: 0.0111 - val_acc: 0.9962\nEpoch 14/50\n - 18s - loss: 0.0276 - acc: 0.9918 - val_loss: 0.0126 - val_acc: 0.9962\nEpoch 15/50\n - 19s - loss: 0.0256 - acc: 0.9926 - val_loss: 0.0116 - val_acc: 0.9963\n\nEpoch 00015: ReduceLROnPlateau reducing learning rate to 0.0001250000059371814.\nEpoch 16/50\n - 19s - loss: 0.0213 - acc: 0.9938 - val_loss: 0.0098 - val_acc: 0.9968\nEpoch 17/50\n - 18s - loss: 0.0207 - acc: 0.9935 - val_loss: 0.0099 - val_acc: 0.9967\nEpoch 18/50\n - 19s - loss: 0.0194 - acc: 0.9939 - val_loss: 0.0104 - val_acc: 0.9963\n\nEpoch 00018: ReduceLROnPlateau reducing learning rate to 6.25000029685907e-05.\nEpoch 19/50\n - 19s - loss: 0.0190 - acc: 0.9945 - val_loss: 0.0096 - val_acc: 0.9963\nEpoch 20/50\n - 19s - loss: 0.0168 - acc: 0.9951 - val_loss: 0.0098 - val_acc: 0.9965\nEpoch 21/50\n - 19s - loss: 0.0180 - acc: 0.9949 - val_loss: 0.0105 - val_acc: 0.9962\n\nEpoch 00021: ReduceLROnPlateau reducing learning rate to 3.125000148429535e-05.\nEpoch 22/50\n - 19s - loss: 0.0155 - acc: 0.9953 - val_loss: 0.0099 - val_acc: 0.9965\nEpoch 23/50\n - 19s - loss: 0.0168 - acc: 0.9949 - val_loss: 0.0098 - val_acc: 0.9963\n\nEpoch 00023: ReduceLROnPlateau reducing learning rate to 1.5625000742147677e-05.\nEpoch 24/50\n - 19s - loss: 0.0139 - acc: 0.9961 - val_loss: 0.0098 - val_acc: 0.9962\nEpoch 00024: early stopping\n4 : Val loss: 0.00977982840661998 : Val acc: 0.9961904761904762 \n\n\nEpoch 1/50\n - 28s - loss: 0.2876 - acc: 0.9102 - val_loss: 0.0818 - val_acc: 0.9740\nEpoch 2/50\n - 19s - loss: 0.1032 - acc: 0.9679 - val_loss: 0.0526 - val_acc: 0.9843\nEpoch 3/50\n - 19s - loss: 0.0848 - acc: 0.9740 - val_loss: 0.0299 - val_acc: 0.9903\nEpoch 4/50\n - 19s - loss: 0.0709 - acc: 0.9775 - val_loss: 0.0455 - val_acc: 0.9871\nEpoch 5/50\n - 19s - loss: 0.0633 - acc: 0.9802 - val_loss: 0.0314 - val_acc: 0.9902\n\nEpoch 00005: ReduceLROnPlateau reducing learning rate to 0.0005000000237487257.\nEpoch 6/50\n - 19s - loss: 0.0427 - acc: 0.9872 - val_loss: 0.0193 - val_acc: 0.9941\nEpoch 7/50\n - 19s - loss: 0.0407 - acc: 0.9869 - val_loss: 0.0179 - val_acc: 0.9941\nEpoch 8/50\n - 19s - loss: 0.0383 - acc: 0.9886 - val_loss: 0.0286 - val_acc: 0.9911\nEpoch 9/50\n - 19s - loss: 0.0370 - acc: 0.9888 - val_loss: 0.0150 - val_acc: 0.9946\nEpoch 10/50\n - 19s - loss: 0.0381 - acc: 0.9888 - val_loss: 0.0142 - val_acc: 0.9951\nEpoch 11/50\n - 19s - loss: 0.0349 - acc: 0.9891 - val_loss: 0.0120 - val_acc: 0.9968\nEpoch 12/50\n - 19s - loss: 0.0314 - acc: 0.9903 - val_loss: 0.0124 - val_acc: 0.9967\nEpoch 13/50\n - 19s - loss: 0.0316 - acc: 0.9906 - val_loss: 0.0115 - val_acc: 0.9968\nEpoch 14/50\n - 19s - loss: 0.0295 - acc: 0.9913 - val_loss: 0.0199 - val_acc: 0.9937\nEpoch 15/50\n - 19s - loss: 0.0296 - acc: 0.9911 - val_loss: 0.0151 - val_acc: 0.9956\n\nEpoch 00015: ReduceLROnPlateau reducing learning rate to 0.0002500000118743628.\nEpoch 16/50\n - 19s - loss: 0.0205 - acc: 0.9935 - val_loss: 0.0121 - val_acc: 0.9956\nEpoch 17/50\n - 19s - loss: 0.0192 - acc: 0.9939 - val_loss: 0.0136 - val_acc: 0.9957\n\nEpoch 00017: ReduceLROnPlateau reducing learning rate to 0.0001250000059371814.\nEpoch 18/50\n - 19s - loss: 0.0167 - acc: 0.9951 - val_loss: 0.0110 - val_acc: 0.9965\nEpoch 19/50\n - 19s - loss: 0.0160 - acc: 0.9952 - val_loss: 0.0105 - val_acc: 0.9970\nEpoch 20/50\n - 19s - loss: 0.0153 - acc: 0.9952 - val_loss: 0.0105 - val_acc: 0.9971\nEpoch 21/50\n - 19s - loss: 0.0146 - acc: 0.9956 - val_loss: 0.0104 - val_acc: 0.9976\n\nEpoch 00021: ReduceLROnPlateau reducing learning rate to 6.25000029685907e-05.\nEpoch 22/50\n",
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": " - 19s - loss: 0.0137 - acc: 0.9958 - val_loss: 0.0106 - val_acc: 0.9971\nEpoch 23/50\n - 19s - loss: 0.0129 - acc: 0.9962 - val_loss: 0.0102 - val_acc: 0.9968\nEpoch 24/50\n - 19s - loss: 0.0123 - acc: 0.9964 - val_loss: 0.0092 - val_acc: 0.9975\nEpoch 25/50\n - 19s - loss: 0.0133 - acc: 0.9960 - val_loss: 0.0098 - val_acc: 0.9970\nEpoch 26/50\n - 19s - loss: 0.0124 - acc: 0.9961 - val_loss: 0.0100 - val_acc: 0.9973\n\nEpoch 00026: ReduceLROnPlateau reducing learning rate to 3.125000148429535e-05.\nEpoch 27/50\n - 19s - loss: 0.0119 - acc: 0.9961 - val_loss: 0.0093 - val_acc: 0.9971\nEpoch 28/50\n - 19s - loss: 0.0111 - acc: 0.9968 - val_loss: 0.0094 - val_acc: 0.9970\n\nEpoch 00028: ReduceLROnPlateau reducing learning rate to 1.5625000742147677e-05.\nEpoch 29/50\n - 19s - loss: 0.0105 - acc: 0.9972 - val_loss: 0.0093 - val_acc: 0.9971\nEpoch 00029: early stopping\n5 : Val loss: 0.009320736908689252 : Val acc: 0.9971428571428571 \n\n\nEpoch 1/50\n - 29s - loss: 0.3562 - acc: 0.8888 - val_loss: 0.0545 - val_acc: 0.9838\nEpoch 2/50\n - 19s - loss: 0.1218 - acc: 0.9628 - val_loss: 0.0498 - val_acc: 0.9849\nEpoch 3/50\n - 19s - loss: 0.0953 - acc: 0.9709 - val_loss: 0.0325 - val_acc: 0.9906\nEpoch 4/50\n - 19s - loss: 0.0813 - acc: 0.9739 - val_loss: 0.0296 - val_acc: 0.9902\nEpoch 5/50\n - 19s - loss: 0.0758 - acc: 0.9772 - val_loss: 0.0472 - val_acc: 0.9849\nEpoch 6/50\n - 19s - loss: 0.0671 - acc: 0.9804 - val_loss: 0.0278 - val_acc: 0.9916\nEpoch 7/50\n - 19s - loss: 0.0599 - acc: 0.9821 - val_loss: 0.0282 - val_acc: 0.9919\nEpoch 8/50\n - 19s - loss: 0.0574 - acc: 0.9823 - val_loss: 0.0261 - val_acc: 0.9911\nEpoch 9/50\n - 19s - loss: 0.0512 - acc: 0.9842 - val_loss: 0.0247 - val_acc: 0.9914\nEpoch 10/50\n - 19s - loss: 0.0525 - acc: 0.9840 - val_loss: 0.0190 - val_acc: 0.9938\nEpoch 11/50\n - 19s - loss: 0.0483 - acc: 0.9855 - val_loss: 0.0183 - val_acc: 0.9932\nEpoch 12/50\n - 19s - loss: 0.0433 - acc: 0.9868 - val_loss: 0.0213 - val_acc: 0.9935\nEpoch 13/50\n - 19s - loss: 0.0438 - acc: 0.9868 - val_loss: 0.0157 - val_acc: 0.9951\nEpoch 14/50\n - 19s - loss: 0.0414 - acc: 0.9878 - val_loss: 0.0322 - val_acc: 0.9898\nEpoch 15/50\n - 19s - loss: 0.0413 - acc: 0.9879 - val_loss: 0.0197 - val_acc: 0.9937\n\nEpoch 00015: ReduceLROnPlateau reducing learning rate to 0.0005000000237487257.\nEpoch 16/50\n - 19s - loss: 0.0279 - acc: 0.9909 - val_loss: 0.0118 - val_acc: 0.9963\nEpoch 17/50\n - 19s - loss: 0.0267 - acc: 0.9920 - val_loss: 0.0110 - val_acc: 0.9970\nEpoch 18/50\n - 19s - loss: 0.0265 - acc: 0.9914 - val_loss: 0.0104 - val_acc: 0.9970\nEpoch 19/50\n - 19s - loss: 0.0253 - acc: 0.9926 - val_loss: 0.0119 - val_acc: 0.9963\nEpoch 20/50\n - 19s - loss: 0.0259 - acc: 0.9913 - val_loss: 0.0109 - val_acc: 0.9971\n\nEpoch 00020: ReduceLROnPlateau reducing learning rate to 0.0002500000118743628.\nEpoch 21/50\n - 19s - loss: 0.0211 - acc: 0.9935 - val_loss: 0.0110 - val_acc: 0.9967\nEpoch 22/50\n - 19s - loss: 0.0193 - acc: 0.9945 - val_loss: 0.0102 - val_acc: 0.9970\nEpoch 23/50\n - 19s - loss: 0.0189 - acc: 0.9942 - val_loss: 0.0092 - val_acc: 0.9975\nEpoch 24/50\n - 19s - loss: 0.0177 - acc: 0.9950 - val_loss: 0.0112 - val_acc: 0.9968\nEpoch 25/50\n - 19s - loss: 0.0189 - acc: 0.9945 - val_loss: 0.0091 - val_acc: 0.9973\n\nEpoch 00025: ReduceLROnPlateau reducing learning rate to 0.0001250000059371814.\nEpoch 26/50\n - 19s - loss: 0.0162 - acc: 0.9950 - val_loss: 0.0085 - val_acc: 0.9976\nEpoch 27/50\n - 19s - loss: 0.0155 - acc: 0.9951 - val_loss: 0.0083 - val_acc: 0.9975\nEpoch 28/50\n - 19s - loss: 0.0167 - acc: 0.9952 - val_loss: 0.0094 - val_acc: 0.9976\nEpoch 29/50\n - 19s - loss: 0.0140 - acc: 0.9956 - val_loss: 0.0093 - val_acc: 0.9975\n\nEpoch 00029: ReduceLROnPlateau reducing learning rate to 6.25000029685907e-05.\nEpoch 30/50\n - 19s - loss: 0.0142 - acc: 0.9954 - val_loss: 0.0087 - val_acc: 0.9973\nEpoch 31/50\n - 19s - loss: 0.0148 - acc: 0.9956 - val_loss: 0.0084 - val_acc: 0.9978\n\nEpoch 00031: ReduceLROnPlateau reducing learning rate to 3.125000148429535e-05.\nEpoch 32/50\n - 19s - loss: 0.0133 - acc: 0.9959 - val_loss: 0.0086 - val_acc: 0.9978\nEpoch 00032: early stopping\n6 : Val loss: 0.008627809460578848 : Val acc: 0.9977777777777778 \n\n\nEpoch 1/50\n - 29s - loss: 0.4515 - acc: 0.8600 - val_loss: 0.0639 - val_acc: 0.9797\nEpoch 2/50\n - 19s - loss: 0.1433 - acc: 0.9559 - val_loss: 0.0597 - val_acc: 0.9811\nEpoch 3/50\n - 19s - loss: 0.1155 - acc: 0.9651 - val_loss: 0.0468 - val_acc: 0.9862\nEpoch 4/50\n - 19s - loss: 0.0931 - acc: 0.9715 - val_loss: 0.0288 - val_acc: 0.9905\nEpoch 5/50\n - 19s - loss: 0.0902 - acc: 0.9730 - val_loss: 0.0388 - val_acc: 0.9883\nEpoch 6/50\n - 19s - loss: 0.0776 - acc: 0.9762 - val_loss: 0.0326 - val_acc: 0.9903\n\nEpoch 00006: ReduceLROnPlateau reducing learning rate to 0.0005000000237487257.\nEpoch 7/50\n - 19s - loss: 0.0581 - acc: 0.9824 - val_loss: 0.0185 - val_acc: 0.9940\nEpoch 8/50\n - 19s - loss: 0.0525 - acc: 0.9840 - val_loss: 0.0227 - val_acc: 0.9932\nEpoch 9/50\n - 19s - loss: 0.0487 - acc: 0.9849 - val_loss: 0.0171 - val_acc: 0.9954\nEpoch 10/50\n - 19s - loss: 0.0490 - acc: 0.9851 - val_loss: 0.0168 - val_acc: 0.9948\nEpoch 11/50\n - 19s - loss: 0.0443 - acc: 0.9861 - val_loss: 0.0155 - val_acc: 0.9952\nEpoch 12/50\n - 19s - loss: 0.0439 - acc: 0.9868 - val_loss: 0.0173 - val_acc: 0.9943\nEpoch 13/50\n - 19s - loss: 0.0409 - acc: 0.9881 - val_loss: 0.0142 - val_acc: 0.9951\nEpoch 14/50\n - 19s - loss: 0.0415 - acc: 0.9878 - val_loss: 0.0206 - val_acc: 0.9937\nEpoch 15/50\n - 19s - loss: 0.0403 - acc: 0.9877 - val_loss: 0.0168 - val_acc: 0.9948\n\nEpoch 00015: ReduceLROnPlateau reducing learning rate to 0.0002500000118743628.\nEpoch 16/50\n - 19s - loss: 0.0315 - acc: 0.9904 - val_loss: 0.0167 - val_acc: 0.9949\nEpoch 17/50\n - 19s - loss: 0.0306 - acc: 0.9911 - val_loss: 0.0135 - val_acc: 0.9957\nEpoch 18/50\n - 19s - loss: 0.0292 - acc: 0.9913 - val_loss: 0.0115 - val_acc: 0.9971\nEpoch 19/50\n - 19s - loss: 0.0288 - acc: 0.9909 - val_loss: 0.0113 - val_acc: 0.9970\nEpoch 20/50\n - 19s - loss: 0.0277 - acc: 0.9918 - val_loss: 0.0123 - val_acc: 0.9967\nEpoch 21/50\n - 19s - loss: 0.0275 - acc: 0.9921 - val_loss: 0.0128 - val_acc: 0.9957\n\nEpoch 00021: ReduceLROnPlateau reducing learning rate to 0.0001250000059371814.\nEpoch 22/50\n - 19s - loss: 0.0244 - acc: 0.9926 - val_loss: 0.0103 - val_acc: 0.9973\nEpoch 23/50\n - 19s - loss: 0.0243 - acc: 0.9923 - val_loss: 0.0107 - val_acc: 0.9970\nEpoch 24/50\n - 19s - loss: 0.0234 - acc: 0.9932 - val_loss: 0.0108 - val_acc: 0.9975\n\nEpoch 00024: ReduceLROnPlateau reducing learning rate to 6.25000029685907e-05.\nEpoch 25/50\n - 19s - loss: 0.0228 - acc: 0.9932 - val_loss: 0.0103 - val_acc: 0.9973\nEpoch 26/50\n - 19s - loss: 0.0213 - acc: 0.9936 - val_loss: 0.0103 - val_acc: 0.9976\n\nEpoch 00026: ReduceLROnPlateau reducing learning rate to 3.125000148429535e-05.\nEpoch 27/50\n - 19s - loss: 0.0201 - acc: 0.9938 - val_loss: 0.0105 - val_acc: 0.9978\nEpoch 28/50\n - 19s - loss: 0.0208 - acc: 0.9940 - val_loss: 0.0104 - val_acc: 0.9975\n\nEpoch 00028: ReduceLROnPlateau reducing learning rate to 1.5625000742147677e-05.\nEpoch 29/50\n - 19s - loss: 0.0187 - acc: 0.9947 - val_loss: 0.0102 - val_acc: 0.9976\nEpoch 30/50\n - 19s - loss: 0.0182 - acc: 0.9948 - val_loss: 0.0101 - val_acc: 0.9975\nEpoch 31/50\n - 19s - loss: 0.0196 - acc: 0.9943 - val_loss: 0.0098 - val_acc: 0.9975\nEpoch 32/50\n - 19s - loss: 0.0197 - acc: 0.9941 - val_loss: 0.0097 - val_acc: 0.9978\nEpoch 33/50\n - 19s - loss: 0.0197 - acc: 0.9943 - val_loss: 0.0095 - val_acc: 0.9975\nEpoch 34/50\n - 19s - loss: 0.0199 - acc: 0.9938 - val_loss: 0.0098 - val_acc: 0.9976\nEpoch 35/50\n - 19s - loss: 0.0187 - acc: 0.9943 - val_loss: 0.0100 - val_acc: 0.9975\n\nEpoch 00035: ReduceLROnPlateau reducing learning rate to 7.812500371073838e-06.\nEpoch 36/50\n - 19s - loss: 0.0190 - acc: 0.9943 - val_loss: 0.0099 - val_acc: 0.9973\nEpoch 37/50\n - 19s - loss: 0.0194 - acc: 0.9939 - val_loss: 0.0098 - val_acc: 0.9975\n\nEpoch 00037: ReduceLROnPlateau reducing learning rate to 3.906250185536919e-06.\nEpoch 38/50\n - 19s - loss: 0.0187 - acc: 0.9946 - val_loss: 0.0098 - val_acc: 0.9975\nEpoch 00038: early stopping\n7 : Val loss: 0.009792417598008696 : Val acc: 0.9974603174603175 \n\n\nEpoch 1/50\n - 30s - loss: 0.3135 - acc: 0.9021 - val_loss: 0.0672 - val_acc: 0.9770\nEpoch 2/50\n",
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": " - 19s - loss: 0.1065 - acc: 0.9667 - val_loss: 0.0502 - val_acc: 0.9852\nEpoch 3/50\n - 19s - loss: 0.0857 - acc: 0.9739 - val_loss: 0.0278 - val_acc: 0.9903\nEpoch 4/50\n - 19s - loss: 0.0730 - acc: 0.9772 - val_loss: 0.0342 - val_acc: 0.9892\nEpoch 5/50\n - 19s - loss: 0.0675 - acc: 0.9798 - val_loss: 0.0430 - val_acc: 0.9868\n\nEpoch 00005: ReduceLROnPlateau reducing learning rate to 0.0005000000237487257.\nEpoch 6/50\n - 19s - loss: 0.0464 - acc: 0.9859 - val_loss: 0.0239 - val_acc: 0.9927\nEpoch 7/50\n - 19s - loss: 0.0412 - acc: 0.9881 - val_loss: 0.0228 - val_acc: 0.9929\nEpoch 8/50\n - 19s - loss: 0.0407 - acc: 0.9876 - val_loss: 0.0280 - val_acc: 0.9929\nEpoch 9/50\n - 19s - loss: 0.0374 - acc: 0.9889 - val_loss: 0.0161 - val_acc: 0.9943\nEpoch 10/50\n - 19s - loss: 0.0382 - acc: 0.9891 - val_loss: 0.0161 - val_acc: 0.9956\nEpoch 11/50\n - 19s - loss: 0.0357 - acc: 0.9895 - val_loss: 0.0122 - val_acc: 0.9970\nEpoch 12/50\n - 19s - loss: 0.0324 - acc: 0.9900 - val_loss: 0.0154 - val_acc: 0.9948\nEpoch 13/50\n - 19s - loss: 0.0324 - acc: 0.9904 - val_loss: 0.0158 - val_acc: 0.9948\n\nEpoch 00013: ReduceLROnPlateau reducing learning rate to 0.0002500000118743628.\nEpoch 14/50\n - 19s - loss: 0.0241 - acc: 0.9931 - val_loss: 0.0141 - val_acc: 0.9951\nEpoch 15/50\n - 19s - loss: 0.0223 - acc: 0.9936 - val_loss: 0.0119 - val_acc: 0.9965\nEpoch 16/50\n - 19s - loss: 0.0218 - acc: 0.9929 - val_loss: 0.0127 - val_acc: 0.9957\nEpoch 17/50\n - 19s - loss: 0.0202 - acc: 0.9936 - val_loss: 0.0101 - val_acc: 0.9971\nEpoch 18/50\n - 19s - loss: 0.0194 - acc: 0.9939 - val_loss: 0.0120 - val_acc: 0.9968\nEpoch 19/50\n - 19s - loss: 0.0197 - acc: 0.9940 - val_loss: 0.0100 - val_acc: 0.9971\n\nEpoch 00019: ReduceLROnPlateau reducing learning rate to 0.0001250000059371814.\nEpoch 20/50\n - 19s - loss: 0.0164 - acc: 0.9952 - val_loss: 0.0117 - val_acc: 0.9970\nEpoch 21/50\n - 19s - loss: 0.0140 - acc: 0.9956 - val_loss: 0.0127 - val_acc: 0.9965\n\nEpoch 00021: ReduceLROnPlateau reducing learning rate to 6.25000029685907e-05.\nEpoch 22/50\n - 19s - loss: 0.0142 - acc: 0.9958 - val_loss: 0.0114 - val_acc: 0.9965\nEpoch 23/50\n - 19s - loss: 0.0132 - acc: 0.9960 - val_loss: 0.0102 - val_acc: 0.9968\n\nEpoch 00023: ReduceLROnPlateau reducing learning rate to 3.125000148429535e-05.\nEpoch 24/50\n - 19s - loss: 0.0112 - acc: 0.9966 - val_loss: 0.0102 - val_acc: 0.9968\nEpoch 00024: early stopping\n8 : Val loss: 0.010241405039458418 : Val acc: 0.9968253968253968 \n\n\nEpoch 1/50\n - 30s - loss: 0.4411 - acc: 0.8630 - val_loss: 0.0663 - val_acc: 0.9797\nEpoch 2/50\n - 20s - loss: 0.1442 - acc: 0.9555 - val_loss: 0.0741 - val_acc: 0.9789\nEpoch 3/50\n - 20s - loss: 0.1114 - acc: 0.9659 - val_loss: 0.0439 - val_acc: 0.9867\nEpoch 4/50\n - 20s - loss: 0.0914 - acc: 0.9723 - val_loss: 0.0504 - val_acc: 0.9848\nEpoch 5/50\n - 19s - loss: 0.0884 - acc: 0.9736 - val_loss: 0.0353 - val_acc: 0.9895\nEpoch 6/50\n - 20s - loss: 0.0810 - acc: 0.9754 - val_loss: 0.0303 - val_acc: 0.9906\nEpoch 7/50\n - 20s - loss: 0.0712 - acc: 0.9784 - val_loss: 0.0249 - val_acc: 0.9938\nEpoch 8/50\n - 20s - loss: 0.0661 - acc: 0.9799 - val_loss: 0.0208 - val_acc: 0.9935\nEpoch 9/50\n - 20s - loss: 0.0605 - acc: 0.9816 - val_loss: 0.0234 - val_acc: 0.9933\nEpoch 10/50\n - 19s - loss: 0.0602 - acc: 0.9818 - val_loss: 0.0161 - val_acc: 0.9935\nEpoch 11/50\n - 20s - loss: 0.0554 - acc: 0.9838 - val_loss: 0.0183 - val_acc: 0.9943\nEpoch 12/50\n - 20s - loss: 0.0503 - acc: 0.9851 - val_loss: 0.0201 - val_acc: 0.9935\n\nEpoch 00012: ReduceLROnPlateau reducing learning rate to 0.0005000000237487257.\nEpoch 13/50\n - 19s - loss: 0.0399 - acc: 0.9879 - val_loss: 0.0102 - val_acc: 0.9965\nEpoch 14/50\n - 20s - loss: 0.0389 - acc: 0.9884 - val_loss: 0.0129 - val_acc: 0.9948\nEpoch 15/50\n - 20s - loss: 0.0368 - acc: 0.9893 - val_loss: 0.0129 - val_acc: 0.9959\n\nEpoch 00015: ReduceLROnPlateau reducing learning rate to 0.0002500000118743628.\nEpoch 16/50\n - 20s - loss: 0.0305 - acc: 0.9902 - val_loss: 0.0090 - val_acc: 0.9963\nEpoch 17/50\n - 19s - loss: 0.0303 - acc: 0.9915 - val_loss: 0.0103 - val_acc: 0.9962\nEpoch 18/50\n - 20s - loss: 0.0284 - acc: 0.9915 - val_loss: 0.0093 - val_acc: 0.9968\n\nEpoch 00018: ReduceLROnPlateau reducing learning rate to 0.0001250000059371814.\nEpoch 19/50\n - 20s - loss: 0.0277 - acc: 0.9917 - val_loss: 0.0090 - val_acc: 0.9970\nEpoch 20/50\n - 20s - loss: 0.0237 - acc: 0.9930 - val_loss: 0.0091 - val_acc: 0.9970\n\nEpoch 00020: ReduceLROnPlateau reducing learning rate to 6.25000029685907e-05.\nEpoch 21/50\n - 20s - loss: 0.0235 - acc: 0.9927 - val_loss: 0.0085 - val_acc: 0.9976\nEpoch 22/50\n - 20s - loss: 0.0220 - acc: 0.9932 - val_loss: 0.0083 - val_acc: 0.9973\nEpoch 23/50\n - 20s - loss: 0.0221 - acc: 0.9936 - val_loss: 0.0083 - val_acc: 0.9976\nEpoch 24/50\n - 20s - loss: 0.0225 - acc: 0.9934 - val_loss: 0.0086 - val_acc: 0.9973\n\nEpoch 00024: ReduceLROnPlateau reducing learning rate to 3.125000148429535e-05.\nEpoch 25/50\n - 20s - loss: 0.0221 - acc: 0.9933 - val_loss: 0.0085 - val_acc: 0.9971\nEpoch 26/50\n - 19s - loss: 0.0221 - acc: 0.9934 - val_loss: 0.0090 - val_acc: 0.9971\n\nEpoch 00026: ReduceLROnPlateau reducing learning rate to 1.5625000742147677e-05.\nEpoch 27/50\n - 20s - loss: 0.0202 - acc: 0.9939 - val_loss: 0.0089 - val_acc: 0.9973\nEpoch 28/50\n - 20s - loss: 0.0218 - acc: 0.9936 - val_loss: 0.0089 - val_acc: 0.9975\n\nEpoch 00028: ReduceLROnPlateau reducing learning rate to 7.812500371073838e-06.\nEpoch 00028: early stopping\n9 : Val loss: 0.008867082067853993 : Val acc: 0.9974603174603175 \n\n\n",
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "execution_count": 101,
          "data": {
            "text/plain": "{'Dropout_0': 0.3272449405085721,\n 'Dropout_1': 0.2919607928329454,\n 'Dropout_2': 0.20616325689743176}"
          },
          "metadata": {}
        }
      ]
    },
    {
      "metadata": {},
      "cell_type": "markdown",
      "source": "## Best Hyperparameters:"
    },
    {
      "metadata": {
        "ExecuteTime": {
          "start_time": "2018-11-20T12:47:30.148348Z",
          "end_time": "2018-11-20T12:47:30.151318Z"
        },
        "trusted": true
      },
      "cell_type": "code",
      "source": "space_eval(params, best)",
      "execution_count": 102,
      "outputs": [
        {
          "output_type": "execute_result",
          "execution_count": 102,
          "data": {
            "text/plain": "{'Dropout_0': 0.3272449405085721,\n 'Dropout_1': 0.2919607928329454,\n 'Dropout_2': 0.20616325689743176}"
          },
          "metadata": {}
        }
      ]
    },
    {
      "metadata": {},
      "cell_type": "markdown",
      "source": "## Best Result"
    },
    {
      "metadata": {
        "ExecuteTime": {
          "start_time": "2018-11-20T12:47:30.152297Z",
          "end_time": "2018-11-20T12:47:30.169070Z"
        },
        "trusted": true
      },
      "cell_type": "code",
      "source": "trials.best_trial['result']",
      "execution_count": 103,
      "outputs": [
        {
          "output_type": "execute_result",
          "execution_count": 103,
          "data": {
            "text/plain": "{'hist': <keras.callbacks.History at 0x7f55181e7ef0>,\n 'loss': -0.9977777777777778,\n 'model': <keras.engine.sequential.Sequential at 0x7f555d038f98>,\n 'status': 'ok'}"
          },
          "metadata": {}
        }
      ]
    },
    {
      "metadata": {
        "ExecuteTime": {
          "end_time": "2018-11-10T05:18:12.461129Z",
          "start_time": "2018-11-10T05:18:12.454484Z"
        }
      },
      "cell_type": "markdown",
      "source": "## Best model:"
    },
    {
      "metadata": {
        "ExecuteTime": {
          "start_time": "2018-11-20T12:47:30.170460Z",
          "end_time": "2018-11-20T12:47:30.181957Z"
        },
        "trusted": true
      },
      "cell_type": "code",
      "source": "best_model = trials.best_trial['result']['model']\n#best_model.save('hyperopt_mnist_best_model.hdf5')",
      "execution_count": 104,
      "outputs": []
    },
    {
      "metadata": {},
      "cell_type": "markdown",
      "source": "## Evaluation on Test data:"
    },
    {
      "metadata": {
        "ExecuteTime": {
          "start_time": "2018-11-20T12:47:30.183377Z",
          "end_time": "2018-11-20T12:47:32.158402Z"
        },
        "trusted": true
      },
      "cell_type": "code",
      "source": "score = best_model.evaluate(x_test, y_test, verbose=1)\nscore",
      "execution_count": 105,
      "outputs": [
        {
          "output_type": "stream",
          "text": "6300/6300 [==============================] - 2s 313us/step\n",
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "execution_count": 105,
          "data": {
            "text/plain": "[0.008627809460578848, 0.9977777777777778]"
          },
          "metadata": {}
        }
      ]
    },
    {
      "metadata": {},
      "cell_type": "markdown",
      "source": "## Prediction for the Submission:"
    },
    {
      "metadata": {
        "ExecuteTime": {
          "start_time": "2018-11-20T12:47:32.160308Z",
          "end_time": "2018-11-20T12:47:45.595749Z"
        },
        "trusted": true
      },
      "cell_type": "code",
      "source": "test = pd.read_csv('../test.csv')\ntest_index = test.index\ntest = test.values.reshape(-1, 28, 28, 1).astype('float32') / 255.0\n\npred = best_model.predict(test, verbose=1)\nresult = pred.argmax(axis=1)",
      "execution_count": 106,
      "outputs": [
        {
          "output_type": "stream",
          "text": "28000/28000 [==============================] - 12s 434us/step\n",
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {},
      "cell_type": "markdown",
      "source": "## Output the Submission csv file:"
    },
    {
      "metadata": {
        "ExecuteTime": {
          "start_time": "2018-11-20T12:47:45.596994Z",
          "end_time": "2018-11-20T12:47:45.652002Z"
        },
        "trusted": true
      },
      "cell_type": "code",
      "source": "submission = pd.DataFrame({'ImageId': test_index+1, 'Label': result})\nsubmission.to_csv('hyperopt_gen_sim.csv', index=False)",
      "execution_count": 107,
      "outputs": []
    },
    {
      "metadata": {},
      "cell_type": "markdown",
      "source": "#### Comparison with the previous result scored at 0.99700"
    },
    {
      "metadata": {
        "ExecuteTime": {
          "start_time": "2018-11-20T12:47:45.653163Z",
          "end_time": "2018-11-20T12:47:45.668368Z"
        },
        "trusted": true
      },
      "cell_type": "code",
      "source": "prev_result = pd.read_csv('hyperopt_gen_sim_99700.csv', index_col=0)\ncurr_result = pd.read_csv('hyperopt_gen_sim.csv', index_col=0)\nmatch_num = np.sum(prev_result.Label.values == curr_result.Label.values)\ncomp_acc = match_num / len(curr_result)\nprint('Approx. accuracy: {0:.5f}'.format(comp_acc))",
      "execution_count": 108,
      "outputs": [
        {
          "output_type": "stream",
          "text": "Approx. accuracy: 0.99750\n",
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "ExecuteTime": {
          "start_time": "2018-11-20T12:47:45.669520Z",
          "end_time": "2018-11-20T12:47:47.998631Z"
        },
        "trusted": true
      },
      "cell_type": "code",
      "source": "idx = prev_result[prev_result.Label.values != curr_result.Label.values].index\nnum=len(idx)\nprint('Difference:',num)\nprint('Image Title: [previous result, current result, id number]')\n\nfig = plt.figure(figsize=(20, (num//10+1)*2))\nfor n in range(num):\n    plt.subplot(num//10+1, 10, n+1)\n    img = test[idx[n]-1].reshape(28,28)\n    plt.imshow(img, cmap='gray')\n    plt.title([prev_result.loc[idx[n]].Label, curr_result.loc[idx[n]].Label, idx[n]])\n    plt.axis('off')\nplt.show()",
      "execution_count": 109,
      "outputs": [
        {
          "output_type": "stream",
          "text": "Difference: 70\nImage Title: [previous result, current result, id number]\n",
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": "<matplotlib.figure.Figure at 0x7f54e3133c88>",
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    }
  ],
  "metadata": {
    "_draft": {
      "nbviewer_url": "https://gist.github.com/48b63f435199da3b3f399063b53b7f06"
    },
    "gist": {
      "id": "48b63f435199da3b3f399063b53b7f06",
      "data": {
        "description": "Kaggle: Digital Recognizer(MNIST) by Hyperopt + fit_generator",
        "public": true
      }
    },
    "kernelspec": {
      "name": "py36",
      "display_name": "py36",
      "language": "python"
    },
    "language_info": {
      "name": "python",
      "version": "3.6.4",
      "mimetype": "text/x-python",
      "codemirror_mode": {
        "name": "ipython",
        "version": 3
      },
      "pygments_lexer": "ipython3",
      "nbconvert_exporter": "python",
      "file_extension": ".py"
    }
  },
  "nbformat": 4,
  "nbformat_minor": 2
}