In [1]:

    

# author - Richard Liao
# Dec 26 2016
# https://richliao.github.io/supervised/classification/2016/11/26/textclassifier-convolutional/
import numpy as np
import pandas as pd
import pickle
from collections import defaultdict
import re

from bs4 import BeautifulSoup

import sys
import os

os.environ['KERAS_BACKEND']='theano'

from keras.preprocessing.text import Tokenizer
from keras.preprocessing.sequence import pad_sequences
from keras.utils.np_utils import to_categorical

from keras.layers import Embedding
from keras.layers import Dense, Input, Flatten
from keras.models import Sequential
from keras.layers.core import Activation
from keras.layers import Conv1D, MaxPooling1D, Embedding, Merge, Dropout, Reshape
from keras.models import Model
from keras.models import load_model
from keras.callbacks import EarlyStopping, ModelCheckpoint
from keras.regularizers import l2

import matplotlib.pyplot as plt
from sklearn import preprocessing
from sklearn.model_selection import train_test_split


# fix random seed for reproducibility
seed = 42
np.random.seed(seed)

%matplotlib inline


    

    




Using Theano backend.
WARNING (theano.tensor.blas): Using NumPy C-API based implementation for BLAS functions.

In [2]:

    

MAX_SEQUENCE_LENGTH = 1000 # top 30
MAX_NB_WORDS = 20000 # more than vocab size
EMBEDDING_DIM = 100
VALIDATION_SPLIT = 0.2


data_train = pd.read_csv('../result_all_windows_labels.csv')

# Cleanup - remove no labels
data_train = data_train[data_train['label'].notnull()]
data_train = data_train[data_train.label != 'environmental']
data_train = data_train[data_train.label != 'religious']
data_train = data_train[data_train.label != 'economical']


label_cat = {'violence/terrorism' : 1, 'misc': 2, 'political': 3, 
#              'religious': 4, 'economical': 5, 'environmental': 6
            }
print(label_cat) 


def to_category(x):
    return label_cat[x]

data_train['target'] = data_train.apply(lambda row: to_category(row['label']), axis=1)


data_train['target'].plot.hist(alpha=0.5)


texts = []
# Get corpus by joining all keywords
for index, row in data_train.iloc[ :, 2:32].iterrows():
    texts.append(u' '.join(row.tolist()))
    
data_train['topicFlat'] = texts

# texts = datdata_traina_train['topicFlat']

labels = data_train['target']

# print(labels)
data_train['topicFlat'].head()


    

    




{'violence/terrorism': 1, 'political': 3, 'misc': 2}






    
Out[2]:





0    syrian assad say syria killed damascus people ...
1    use osc copyrighted_material dissemination usa...
2    will year can people one country party make sa...
3    quot apos say the we it reuters terrorists ass...
4    baghdad iraq sunni killed bomb iraqi attacks w...
Name: topicFlat, dtype: object






    

In [3]:

    

%%time
    

tokenizer = Tokenizer(num_words=MAX_NB_WORDS)
tokenizer.fit_on_texts(texts)


# max_fatures = 10000
# tokenizer = Tokenizer(num_words=max_fatures, split=' ')
# tokenizer.fit_on_texts(data_train['topicFlat'].values)
# X = tokenizer.texts_to_sequences(data_train['topicFlat'].values)
# X = pad_sequences(X)

# Y = pd.get_dummies(data_train['label']).values
# x_train, x_test, y_train, y_test = train_test_split(X, Y, test_size = 0.33, random_state = 42)

# print(y_train[100])
# print(x_train.shape,y_train.shape)
# print(x_test.shape, y_test.shape)

# print('Number of positive and negative reviews in traing and validation set ')
# print (y_train.sum(axis=0))
# print (y_test.sum(axis=0))

# print(X.shape[0])

sequences = tokenizer.texts_to_sequences(texts)

# print(sequences[0])
# print (max(sequences, key=len))

word_index = tokenizer.word_index
print('Found %s unique tokens.' % len(word_index)) # all the tokens in corpus
# # print(word_index)

data = pad_sequences(sequences, maxlen=MAX_SEQUENCE_LENGTH)
# print(data[0])

labels = to_categorical(np.asarray(labels))
print('Shape of data tensor:', data.shape)
print('Shape of label tensor:', labels.shape)


    

    




Found 6077 unique tokens.
Shape of data tensor: (1449, 1000)
Shape of label tensor: (1449, 4)
CPU times: user 84 ms, sys: 0 ns, total: 84 ms
Wall time: 82.9 ms

In [4]:

    

indices = np.arange(data.shape[0])
np.random.shuffle(indices)
data = data[indices]
labels = labels[indices]
nb_validation_samples = int(VALIDATION_SPLIT * data.shape[0])

x_train = data[:-nb_validation_samples]
y_train = labels[:-nb_validation_samples]
x_test = data[-nb_validation_samples:]
y_test = labels[-nb_validation_samples:]


print('Number of positive and negative reviews in traing and validation set ')
print (y_train.sum(axis=0))
print (y_test.sum(axis=0))


    

    




Number of positive and negative reviews in traing and validation set 
[   0.  466.  290.  404.]
[   0.  109.   67.  113.]

In [5]:

    

%%time

# https://www.kaggle.com/rtatman/glove-global-vectors-for-word-representation/data

GLOVE_DIR = "../data/"
embeddings_index = {}
f = open(os.path.join(GLOVE_DIR, 'glove.6B.100d.txt'))
for line in f:
    values = line.split()
    word = values[0]
    coefs = np.asarray(values[1:], dtype='float32')
    embeddings_index[word] = coefs
f.close()

print('Total %s word vectors in Glove 6B 100d.' % len(embeddings_index))


    

    




Total 400000 word vectors in Glove 6B 100d.
CPU times: user 8.12 s, sys: 160 ms, total: 8.28 s
Wall time: 8.28 s

In [6]:

    

def plot_history(history):
    # list all data in history
    print(history.history.keys())
    # summarize history for accuracy
    plt.plot(history.history['acc'])
    plt.plot(history.history['val_acc'])
    plt.title('model accuracy')
    plt.ylabel('accuracy')
    plt.xlabel('epoch')
    plt.legend(['train', 'test'], loc='upper left')
    plt.show()
    # summarize history for loss
    plt.plot(history.history['loss'])
    plt.plot(history.history['val_loss'])
    plt.title('model loss')
    plt.ylabel('loss')
    plt.xlabel('epoch')
    plt.legend(['train', 'test'], loc='upper left')
    plt.show()


    

In [7]:

    

%%time


embedding_matrix = np.random.random((len(word_index) + 1, EMBEDDING_DIM))
for word, i in word_index.items():
    embedding_vector = embeddings_index.get(word)
    if embedding_vector is not None:
        # words not found in embedding index will be all-zeros.
        embedding_matrix[i] = embedding_vector
        
embedding_layer = Embedding(len(word_index) + 1,
                            EMBEDDING_DIM,
                            weights=[embedding_matrix],
#                             input_length=MAX_SEQUENCE_LENGTH,
                            trainable=True)


# Embedding(5000, 32, input_length=500)


    

    




CPU times: user 12 ms, sys: 0 ns, total: 12 ms
Wall time: 12.6 ms

In [8]:

    

%%time

epochs=20

sequence_input = Input(shape=(MAX_SEQUENCE_LENGTH,), dtype='int32')
embedded_sequences = embedding_layer(sequence_input)
l_cov1= Conv1D(128, 5, activation='relu')(embedded_sequences)
l_pool1 = MaxPooling1D(5)(l_cov1)
l_cov2 = Conv1D(128, 5, activation='relu')(l_pool1)
l_pool2 = MaxPooling1D(5)(l_cov2)
l_cov3 = Conv1D(128, 5, activation='relu')(l_pool2)
l_pool3 = MaxPooling1D(35)(l_cov3)  # global max pooling
l_flat = Flatten()(l_pool3)
l_dense = Dense(128, activation='relu')(l_flat)
preds = Dense(len(label_cat) + 1, activation='softmax')(l_dense)

model = Model(sequence_input, preds)
model.compile(loss='categorical_crossentropy',
              optimizer='adam', metrics=['accuracy'])

print("model fitting - simplified convolutional neural network")
model.summary()

model_name = 'topicConvNet-Reg.h5'
early_stop = EarlyStopping(monitor='val_loss', patience=4, verbose=1)
# best_model = ModelCheckpoint(model_name, verbose=1, save_best_only=True)

best_model = ModelCheckpoint('weights.{epoch:03d}-{val_acc:.4f}.hdf5', 
                             monitor='val_acc', verbose=1, save_best_only=True, mode='auto')

network_hist = model.fit(x_train, y_train, validation_data=(x_test, y_test), 
          verbose=True, callbacks=[early_stop, best_model], validation_split=0.2,
          epochs=epochs, batch_size=50)

score, acc = model.evaluate(x_test, y_test, batch_size=50)
print('Test score:', score)
print('Test accuracy:', acc)

# print(network_hist.history)
plot_history(network_hist)


    

    




model fitting - simplified convolutional neural network
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
input_1 (InputLayer)         (None, 1000)              0         
_________________________________________________________________
embedding_1 (Embedding)      (None, 1000, 100)         607800    
_________________________________________________________________
conv1d_1 (Conv1D)            (None, 996, 128)          64128     
_________________________________________________________________
max_pooling1d_1 (MaxPooling1 (None, 199, 128)          0         
_________________________________________________________________
conv1d_2 (Conv1D)            (None, 195, 128)          82048     
_________________________________________________________________
max_pooling1d_2 (MaxPooling1 (None, 39, 128)           0         
_________________________________________________________________
conv1d_3 (Conv1D)            (None, 35, 128)           82048     
_________________________________________________________________
max_pooling1d_3 (MaxPooling1 (None, 1, 128)            0         
_________________________________________________________________
flatten_1 (Flatten)          (None, 128)               0         
_________________________________________________________________
dense_1 (Dense)              (None, 128)               16512     
_________________________________________________________________
dense_2 (Dense)              (None, 4)                 516       
=================================================================
Total params: 853,052
Trainable params: 853,052
Non-trainable params: 0
_________________________________________________________________
Train on 1160 samples, validate on 289 samples
Epoch 1/20
1150/1160 [============================>.] - ETA: 0s - loss: 0.9545 - acc: 0.5183Epoch 00001: val_acc improved from -inf to 0.77509, saving model to weights.001-0.7751.hdf5
1160/1160 [==============================] - 39s 33ms/step - loss: 0.9521 - acc: 0.5172 - val_loss: 0.6454 - val_acc: 0.7751
Epoch 2/20
1150/1160 [============================>.] - ETA: 0s - loss: 0.5879 - acc: 0.7296Epoch 00002: val_acc did not improve
1160/1160 [==============================] - 39s 34ms/step - loss: 0.5867 - acc: 0.7302 - val_loss: 0.5485 - val_acc: 0.7612
Epoch 3/20
1150/1160 [============================>.] - ETA: 0s - loss: 0.4662 - acc: 0.8235Epoch 00003: val_acc improved from 0.77509 to 0.82699, saving model to weights.003-0.8270.hdf5
1160/1160 [==============================] - 41s 35ms/step - loss: 0.4639 - acc: 0.8250 - val_loss: 0.4702 - val_acc: 0.8270
Epoch 4/20
1150/1160 [============================>.] - ETA: 0s - loss: 0.3525 - acc: 0.8591Epoch 00004: val_acc did not improve
1160/1160 [==============================] - 38s 32ms/step - loss: 0.3568 - acc: 0.8578 - val_loss: 0.5672 - val_acc: 0.7958
Epoch 5/20
1150/1160 [============================>.] - ETA: 0s - loss: 0.3109 - acc: 0.8878Epoch 00005: val_acc improved from 0.82699 to 0.83045, saving model to weights.005-0.8304.hdf5
1160/1160 [==============================] - 39s 34ms/step - loss: 0.3106 - acc: 0.8871 - val_loss: 0.4836 - val_acc: 0.8304
Epoch 6/20
1150/1160 [============================>.] - ETA: 0s - loss: 0.2290 - acc: 0.9209Epoch 00006: val_acc improved from 0.83045 to 0.84429, saving model to weights.006-0.8443.hdf5
1160/1160 [==============================] - 39s 34ms/step - loss: 0.2282 - acc: 0.9207 - val_loss: 0.4833 - val_acc: 0.8443
Epoch 7/20
1150/1160 [============================>.] - ETA: 0s - loss: 0.1812 - acc: 0.9443Epoch 00007: val_acc did not improve
1160/1160 [==============================] - 38s 33ms/step - loss: 0.1802 - acc: 0.9448 - val_loss: 0.5020 - val_acc: 0.8374
Epoch 00007: early stopping
289/289 [==============================] - 2s 7ms/step
Test score: 0.501955389358
Test accuracy: 0.837370237471
dict_keys(['loss', 'val_acc', 'val_loss', 'acc'])






    













    













    




CPU times: user 9min 24s, sys: 24min 53s, total: 34min 17s
Wall time: 4min 40s

In [9]:

    

from IPython.display import SVG
from keras.utils.vis_utils import model_to_dot

SVG(model_to_dot(model).create(prog='dot', format='svg'))


    

    
Out[9]:

In [10]:

    

embedding_matrix = np.random.random((len(word_index) + 1, EMBEDDING_DIM))
for word, i in word_index.items():
    embedding_vector = embeddings_index.get(word)
    if embedding_vector is not None:
        # words not found in embedding index will be all-zeros.
        embedding_matrix[i] = embedding_vector
        
embedding_layer = Embedding(len(word_index) + 1,
                            EMBEDDING_DIM,
                            weights=[embedding_matrix],
                            input_length=MAX_SEQUENCE_LENGTH,
                            trainable=True)

# applying a more complex convolutional approach
convs = []
filter_sizes = [3,4,5]

sequence_input = Input(shape=(MAX_SEQUENCE_LENGTH,), dtype='int32')
embedded_sequences = embedding_layer(sequence_input)

for fsz in filter_sizes:
    l_conv = Conv1D(filters=128,filter_length=fsz, activation='relu')(embedded_sequences)
    l_pool = MaxPooling1D(5)(l_conv)
    convs.append(l_pool)
    
l_merge = Merge(mode='concat', concat_axis=1)(convs)
l_cov1= Conv1D(filters=128, kernel_size=5, activation='relu')(l_merge)
l_pool1 = MaxPooling1D(5)(l_cov1)
l_cov2 = Conv1D(filters=128, kernel_size=5, activation='relu')(l_pool1)
l_pool2 = MaxPooling1D(30)(l_cov2)
l_flat = Flatten()(l_pool2)
l_dense = Dense(128, activation='relu')(l_flat)
preds = Dense(len(label_cat) + 1, activation='softmax')(l_dense)

model = Model(sequence_input, preds)
model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])

print("model fitting - more complex convolutional neural network")
model.summary()

model_name = 'topicConv_20.h5'
early_stop = EarlyStopping(monitor='val_loss', patience=2, verbose=1)
# best_model = ModelCheckpoint(model_name, verbose=0, save_best_only=True)

best_model = ModelCheckpoint('weights.{epoch:03d}-{val_acc:.4f}.hdf5', 
                             monitor='val_acc', verbose=1, save_best_only=True, mode='auto')


network_hist = model.fit(x_train, y_train, validation_data=(x_test, y_test), 
          verbose=True, callbacks=[early_stop, best_model],
          epochs=20, batch_size=50)

score, acc = model.evaluate(x_test, y_test, batch_size=50)
print('Test score:', score)
print('Test accuracy:', acc)

# print(network_hist.history)
plot_history(network_hist)


    

    




/usr/local/lib/python3.5/dist-packages/ipykernel_launcher.py:25: UserWarning: Update your `Conv1D` call to the Keras 2 API: `Conv1D(filters=128, kernel_size=3, activation="relu")`
/usr/local/lib/python3.5/dist-packages/ipykernel_launcher.py:25: UserWarning: Update your `Conv1D` call to the Keras 2 API: `Conv1D(filters=128, kernel_size=4, activation="relu")`
/usr/local/lib/python3.5/dist-packages/ipykernel_launcher.py:25: UserWarning: Update your `Conv1D` call to the Keras 2 API: `Conv1D(filters=128, kernel_size=5, activation="relu")`
/usr/local/lib/python3.5/dist-packages/ipykernel_launcher.py:29: UserWarning: The `Merge` layer is deprecated and will be removed after 08/2017. Use instead layers from `keras.layers.merge`, e.g. `add`, `concatenate`, etc.






    




model fitting - more complex convolutional neural network
__________________________________________________________________________________________________
Layer (type)                    Output Shape         Param #     Connected to                     
==================================================================================================
input_2 (InputLayer)            (None, 1000)         0                                            
__________________________________________________________________________________________________
embedding_2 (Embedding)         (None, 1000, 100)    607800      input_2[0][0]                    
__________________________________________________________________________________________________
conv1d_4 (Conv1D)               (None, 998, 128)     38528       embedding_2[0][0]                
__________________________________________________________________________________________________
conv1d_5 (Conv1D)               (None, 997, 128)     51328       embedding_2[0][0]                
__________________________________________________________________________________________________
conv1d_6 (Conv1D)               (None, 996, 128)     64128       embedding_2[0][0]                
__________________________________________________________________________________________________
max_pooling1d_4 (MaxPooling1D)  (None, 199, 128)     0           conv1d_4[0][0]                   
__________________________________________________________________________________________________
max_pooling1d_5 (MaxPooling1D)  (None, 199, 128)     0           conv1d_5[0][0]                   
__________________________________________________________________________________________________
max_pooling1d_6 (MaxPooling1D)  (None, 199, 128)     0           conv1d_6[0][0]                   
__________________________________________________________________________________________________
merge_1 (Merge)                 (None, 597, 128)     0           max_pooling1d_4[0][0]            
                                                                 max_pooling1d_5[0][0]            
                                                                 max_pooling1d_6[0][0]            
__________________________________________________________________________________________________
conv1d_7 (Conv1D)               (None, 593, 128)     82048       merge_1[0][0]                    
__________________________________________________________________________________________________
max_pooling1d_7 (MaxPooling1D)  (None, 118, 128)     0           conv1d_7[0][0]                   
__________________________________________________________________________________________________
conv1d_8 (Conv1D)               (None, 114, 128)     82048       max_pooling1d_7[0][0]            
__________________________________________________________________________________________________
max_pooling1d_8 (MaxPooling1D)  (None, 3, 128)       0           conv1d_8[0][0]                   
__________________________________________________________________________________________________
flatten_2 (Flatten)             (None, 384)          0           max_pooling1d_8[0][0]            
__________________________________________________________________________________________________
dense_3 (Dense)                 (None, 128)          49280       flatten_2[0][0]                  
__________________________________________________________________________________________________
dense_4 (Dense)                 (None, 4)            516         dense_3[0][0]                    
==================================================================================================
Total params: 975,676
Trainable params: 975,676
Non-trainable params: 0
__________________________________________________________________________________________________
Train on 1160 samples, validate on 289 samples
Epoch 1/20
1150/1160 [============================>.] - ETA: 0s - loss: 1.0777 - acc: 0.4496Epoch 00001: val_acc improved from -inf to 0.63322, saving model to weights.001-0.6332.hdf5
1160/1160 [==============================] - 76s 65ms/step - loss: 1.0758 - acc: 0.4509 - val_loss: 0.8386 - val_acc: 0.6332
Epoch 2/20
1150/1160 [============================>.] - ETA: 0s - loss: 0.6540 - acc: 0.7513Epoch 00002: val_acc improved from 0.63322 to 0.80277, saving model to weights.002-0.8028.hdf5
1160/1160 [==============================] - 77s 66ms/step - loss: 0.6538 - acc: 0.7500 - val_loss: 0.5136 - val_acc: 0.8028
Epoch 3/20
1150/1160 [============================>.] - ETA: 0s - loss: 0.5013 - acc: 0.8096Epoch 00003: val_acc improved from 0.80277 to 0.80969, saving model to weights.003-0.8097.hdf5
1160/1160 [==============================] - 84s 72ms/step - loss: 0.5019 - acc: 0.8086 - val_loss: 0.4972 - val_acc: 0.8097
Epoch 4/20
1150/1160 [============================>.] - ETA: 0s - loss: 0.3532 - acc: 0.8748Epoch 00004: val_acc did not improve
1160/1160 [==============================] - 79s 68ms/step - loss: 0.3522 - acc: 0.8759 - val_loss: 0.5626 - val_acc: 0.7889
Epoch 5/20
1150/1160 [============================>.] - ETA: 0s - loss: 0.3076 - acc: 0.8974Epoch 00005: val_acc improved from 0.80969 to 0.83045, saving model to weights.005-0.8304.hdf5
1160/1160 [==============================] - 76s 66ms/step - loss: 0.3054 - acc: 0.8983 - val_loss: 0.4788 - val_acc: 0.8304
Epoch 6/20
1150/1160 [============================>.] - ETA: 0s - loss: 0.2362 - acc: 0.9148Epoch 00006: val_acc improved from 0.83045 to 0.83737, saving model to weights.006-0.8374.hdf5
1160/1160 [==============================] - 77s 66ms/step - loss: 0.2419 - acc: 0.9121 - val_loss: 0.4716 - val_acc: 0.8374
Epoch 7/20
1150/1160 [============================>.] - ETA: 0s - loss: 0.2083 - acc: 0.9322Epoch 00007: val_acc did not improve
1160/1160 [==============================] - 77s 66ms/step - loss: 0.2070 - acc: 0.9328 - val_loss: 0.5093 - val_acc: 0.8270
Epoch 8/20
1150/1160 [============================>.] - ETA: 0s - loss: 0.1352 - acc: 0.9670Epoch 00008: val_acc did not improve
1160/1160 [==============================] - 81s 70ms/step - loss: 0.1345 - acc: 0.9672 - val_loss: 0.5098 - val_acc: 0.8304
Epoch 00008: early stopping
289/289 [==============================] - 5s 18ms/step
Test score: 0.509825248722
Test accuracy: 0.830449808221
dict_keys(['loss', 'val_acc', 'val_loss', 'acc'])






    













    

In [11]:

    

# EMBEDDING_DIM = 32
# embedding_matrix = np.random.random((len(word_index) + 1, EMBEDDING_DIM))
# for word, i in word_index.items():
#     embedding_vector = embeddings_index.get(word)
#     if embedding_vector is not None:
#         # words not found in embedding index will be all-zeros.
#         embedding_matrix[i] = embedding_vector
        
# embedding_layer = Embedding(len(word_index) + 1,
#                             EMBEDDING_DIM,
#                             weights=[embedding_matrix],
#                             input_length=MAX_SEQUENCE_LENGTH,
#                             trainable=True)
# create the model
model = Sequential()
model.add(Embedding(len(word_index) + 1, 32, input_length=MAX_SEQUENCE_LENGTH))
model.add(Conv1D(filters=32, kernel_size=3, padding='same', activation='relu'))
model.add(MaxPooling1D(pool_size=2))
model.add(Flatten())
model.add(Dense(250, activation='relu'))
model.add(Dense(len(label_cat) + 1, activation='sigmoid'))

model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])
print(model.summary())


model_name = 'topicConv_simple.h5'
early_stop = EarlyStopping(monitor='val_loss', patience=4, verbose=1)
best_model = ModelCheckpoint(model_name, verbose=1, save_best_only=True)

network_hist = model.fit(x_train, y_train, validation_data=(x_test, y_test), 
          verbose=True, callbacks=[early_stop, best_model],
          epochs=20, batch_size=50)

# Evaluation on the test set
score, acc = model.evaluate(x_test, y_test)
print("Test score: %.2f%%" % (score*100))
print("Test accuracy: %.2f%%" % (acc*100))

# print(network_hist.history)
plot_history(network_hist)


    

    




_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
embedding_3 (Embedding)      (None, 1000, 32)          194496    
_________________________________________________________________
conv1d_9 (Conv1D)            (None, 1000, 32)          3104      
_________________________________________________________________
max_pooling1d_9 (MaxPooling1 (None, 500, 32)           0         
_________________________________________________________________
flatten_3 (Flatten)          (None, 16000)             0         
_________________________________________________________________
dense_5 (Dense)              (None, 250)               4000250   
_________________________________________________________________
dense_6 (Dense)              (None, 4)                 1004      
=================================================================
Total params: 4,198,854
Trainable params: 4,198,854
Non-trainable params: 0
_________________________________________________________________
None
Train on 1160 samples, validate on 289 samples
Epoch 1/20
1150/1160 [============================>.] - ETA: 0s - loss: 1.1161 - acc: 0.4017Epoch 00001: val_loss improved from inf to 1.07615, saving model to topicConv_simple.h5
1160/1160 [==============================] - 6s 6ms/step - loss: 1.1160 - acc: 0.4017 - val_loss: 1.0762 - val_acc: 0.3772
Epoch 2/20
1150/1160 [============================>.] - ETA: 0s - loss: 1.0771 - acc: 0.4017Epoch 00002: val_loss improved from 1.07615 to 1.04660, saving model to topicConv_simple.h5
1160/1160 [==============================] - 8s 7ms/step - loss: 1.0769 - acc: 0.4017 - val_loss: 1.0466 - val_acc: 0.3772
Epoch 3/20
1150/1160 [============================>.] - ETA: 0s - loss: 0.9227 - acc: 0.4061Epoch 00003: val_loss improved from 1.04660 to 0.82165, saving model to topicConv_simple.h5
1160/1160 [==============================] - 8s 7ms/step - loss: 0.9220 - acc: 0.4052 - val_loss: 0.8217 - val_acc: 0.4325
Epoch 4/20
1150/1160 [============================>.] - ETA: 0s - loss: 0.6457 - acc: 0.6748Epoch 00004: val_loss improved from 0.82165 to 0.71154, saving model to topicConv_simple.h5
1160/1160 [==============================] - 7s 6ms/step - loss: 0.6434 - acc: 0.6767 - val_loss: 0.7115 - val_acc: 0.6228
Epoch 5/20
1150/1160 [============================>.] - ETA: 0s - loss: 0.4131 - acc: 0.8174Epoch 00005: val_loss improved from 0.71154 to 0.58860, saving model to topicConv_simple.h5
1160/1160 [==============================] - 7s 6ms/step - loss: 0.4138 - acc: 0.8181 - val_loss: 0.5886 - val_acc: 0.7785
Epoch 6/20
1150/1160 [============================>.] - ETA: 0s - loss: 0.2658 - acc: 0.9052Epoch 00006: val_loss did not improve
1160/1160 [==============================] - 7s 6ms/step - loss: 0.2638 - acc: 0.9060 - val_loss: 0.6090 - val_acc: 0.7855
Epoch 7/20
1150/1160 [============================>.] - ETA: 0s - loss: 0.1718 - acc: 0.9443Epoch 00007: val_loss did not improve
1160/1160 [==============================] - 7s 6ms/step - loss: 0.1710 - acc: 0.9448 - val_loss: 0.6487 - val_acc: 0.7820
Epoch 8/20
1150/1160 [============================>.] - ETA: 0s - loss: 0.1126 - acc: 0.9661Epoch 00008: val_loss did not improve
1160/1160 [==============================] - 7s 6ms/step - loss: 0.1123 - acc: 0.9664 - val_loss: 0.6634 - val_acc: 0.7751
Epoch 9/20
1150/1160 [============================>.] - ETA: 0s - loss: 0.0646 - acc: 0.9835Epoch 00009: val_loss did not improve
1160/1160 [==============================] - 7s 6ms/step - loss: 0.0643 - acc: 0.9836 - val_loss: 0.7335 - val_acc: 0.7751
Epoch 00009: early stopping
289/289 [==============================] - 1s 2ms/step
Test score: 73.35%
Test accuracy: 77.51%
dict_keys(['loss', 'val_acc', 'val_loss', 'acc'])