2015-12-09 02:49:14 +00:00
|
|
|
'''Train and evaluate a simple MLP on the Reuters newswire topic classification task.
|
|
|
|
GPU run command:
|
|
|
|
THEANO_FLAGS=mode=FAST_RUN,device=gpu,floatX=float32 python examples/reuters_mlp.py
|
|
|
|
CPU run command:
|
|
|
|
python examples/reuters_mlp.py
|
|
|
|
'''
|
|
|
|
|
2015-04-16 03:18:40 +00:00
|
|
|
from __future__ import print_function
|
2015-03-28 00:59:42 +00:00
|
|
|
import numpy as np
|
2015-07-26 08:00:18 +00:00
|
|
|
np.random.seed(1337) # for reproducibility
|
2015-03-28 00:59:42 +00:00
|
|
|
|
|
|
|
from keras.datasets import reuters
|
|
|
|
from keras.models import Sequential
|
|
|
|
from keras.layers.core import Dense, Dropout, Activation
|
|
|
|
from keras.layers.normalization import BatchNormalization
|
|
|
|
from keras.utils import np_utils
|
|
|
|
from keras.preprocessing.text import Tokenizer
|
|
|
|
|
2015-06-02 03:11:44 +00:00
|
|
|
max_words = 1000
|
2015-05-31 01:46:00 +00:00
|
|
|
batch_size = 32
|
2015-06-19 22:38:33 +00:00
|
|
|
nb_epoch = 5
|
2015-03-28 00:59:42 +00:00
|
|
|
|
2015-12-09 02:49:14 +00:00
|
|
|
print('Loading data...')
|
2015-03-28 00:59:42 +00:00
|
|
|
(X_train, y_train), (X_test, y_test) = reuters.load_data(nb_words=max_words, test_split=0.2)
|
2015-04-16 03:18:40 +00:00
|
|
|
print(len(X_train), 'train sequences')
|
|
|
|
print(len(X_test), 'test sequences')
|
2015-03-28 00:59:42 +00:00
|
|
|
|
|
|
|
nb_classes = np.max(y_train)+1
|
2015-04-16 03:18:40 +00:00
|
|
|
print(nb_classes, 'classes')
|
2015-03-28 00:59:42 +00:00
|
|
|
|
2015-12-09 02:49:14 +00:00
|
|
|
print('Vectorizing sequence data...')
|
2015-03-28 00:59:42 +00:00
|
|
|
tokenizer = Tokenizer(nb_words=max_words)
|
2015-12-09 02:49:14 +00:00
|
|
|
X_train = tokenizer.sequences_to_matrix(X_train, mode='binary')
|
|
|
|
X_test = tokenizer.sequences_to_matrix(X_test, mode='binary')
|
2015-04-16 03:18:40 +00:00
|
|
|
print('X_train shape:', X_train.shape)
|
|
|
|
print('X_test shape:', X_test.shape)
|
2015-03-28 00:59:42 +00:00
|
|
|
|
2015-12-09 02:49:14 +00:00
|
|
|
print('Convert class vector to binary class matrix (for use with categorical_crossentropy)')
|
2015-03-28 00:59:42 +00:00
|
|
|
Y_train = np_utils.to_categorical(y_train, nb_classes)
|
|
|
|
Y_test = np_utils.to_categorical(y_test, nb_classes)
|
2015-04-16 03:18:40 +00:00
|
|
|
print('Y_train shape:', Y_train.shape)
|
|
|
|
print('Y_test shape:', Y_test.shape)
|
2015-03-28 00:59:42 +00:00
|
|
|
|
2015-12-09 02:49:14 +00:00
|
|
|
print('Building model...')
|
2015-03-28 00:59:42 +00:00
|
|
|
model = Sequential()
|
2015-10-05 01:44:49 +00:00
|
|
|
model.add(Dense(512, input_shape=(max_words,)))
|
2015-03-28 00:59:42 +00:00
|
|
|
model.add(Activation('relu'))
|
|
|
|
model.add(Dropout(0.5))
|
2015-10-05 01:44:49 +00:00
|
|
|
model.add(Dense(nb_classes))
|
2015-03-28 00:59:42 +00:00
|
|
|
model.add(Activation('softmax'))
|
|
|
|
|
2015-04-16 03:18:40 +00:00
|
|
|
model.compile(loss='categorical_crossentropy', optimizer='adam')
|
|
|
|
|
2015-06-19 22:38:33 +00:00
|
|
|
history = model.fit(X_train, Y_train, nb_epoch=nb_epoch, batch_size=batch_size, verbose=1, show_accuracy=True, validation_split=0.1)
|
2015-06-12 05:34:23 +00:00
|
|
|
score = model.evaluate(X_test, Y_test, batch_size=batch_size, verbose=1, show_accuracy=True)
|
2015-04-27 21:18:20 +00:00
|
|
|
print('Test score:', score[0])
|
|
|
|
print('Test accuracy:', score[1])
|