def lstm_improved(X_train, y_train, X_test, y_test, vocab_size):
X_train = tf.keras.preprocessing.sequence.pad_sequences(X_train, maxlen=50)
X_test = tf.keras.preprocessing.sequence.pad_sequences(X_test, maxlen=50)
print('X_train shape:', X_train.shape)
print('X_test shape:', X_test.shape)
print('Build model...')
model = tf.keras.Sequential()
model.add(tf.keras.layers.Embedding(vocab_size, EMBED_SIZE, input_length=50))
model.add(tf.keras.layers.SpatialDropout1D(0.2))
peephole_lstm_cells = tfa.rnn.PeepholeLSTMCell(HIDDEN_SIZE)
model.add(tf.keras.layers.RNN(peephole_lstm_cells))
model.add(tf.keras.layers.Dense(1, activation="sigmoid"))
opt = tf.keras.optimizers.SGD(learning_rate=0.001, momentum=0.99, decay=1e-6, nesterov=True)
model.compile(loss=root_mean_squared_logarithmic_error,
optimizer=opt)
# model.compile(optimizer='adam',
# loss='binary_crossentropy')
#
# model.compile(loss='binary_crossentropy',
# optimizer='rmsprop')
model.fit(X_train, y_train, batch_size=BATCH_SIZE,
epochs=EPOCHS,
validation_data=(X_test, y_test))
X_pred = model.predict(X_test)
results = [result[0] for result in X_pred]
readResult(y_test, results, name="LSTM+", form="JSON")
return model
def lstm_zhang(X_train, y_train, X_test, y_test, vocab_size):
X_train = tf.keras.preprocessing.sequence.pad_sequences(X_train, maxlen=50)
X_test = tf.keras.preprocessing.sequence.pad_sequences(X_test, maxlen=50)
print('X_train shape:', X_train.shape)
print('X_test shape:', X_test.shape)
print('Build model...')
model = tf.keras.Sequential()
model.add(tf.keras.layers.Embedding(vocab_size, EMBED_SIZE, input_length=50))
model.add(tf.keras.layers.SpatialDropout1D(0.2))
model.add(tf.keras.layers.LSTM(HIDDEN_SIZE))
model.add(tf.keras.layers.Dense(1, activation="sigmoid"))
model.compile(optimizer='adam',
loss='binary_crossentropy')
model.compile(loss='binary_crossentropy',
optimizer='rmsprop')
model.fit(X_train, y_train, batch_size=BATCH_SIZE,
epochs=EPOCHS,
validation_data=(X_test, y_test), verbose=0)
X_pred = model.predict(X_test)
results = [result[0] for result in X_pred]
readResult(y_test, results, name="LSTM", form="JSON")
return model
def sgd_classify(trains, tests):
trains, tests = word_transform(trains, tests)
model = make_pipeline(StandardScaler(with_mean=False),
SGDClassifier(max_iter=1000, tol=TOLERANCE))
model.fit(trains['tfidf'], trains['target'])
predicted = model.predict(tests['tfidf'])
readResult(tests['target'], predicted, name="SGD", form="JSON")
return model
def svm_classify(trains, tests):
trains, tests = word_transform(trains, tests)
print("trains[0]:")
print(trains['tfidf'])
model = make_pipeline(StandardScaler(with_mean=False),
LinearSVC(tol=TOLERANCE, max_iter=2000000))
model.fit(trains['tfidf'], trains['target'])
predicted = model.predict(tests['tfidf'])
print("predicted: ", predicted)
print("target : ", tests['target'])
readResult(tests['target'], predicted, name="SVM", form="JSON")
return model
def cnn_zhang(X_train, y_train, X_test, y_test, vocab_size):
X_train = tf.keras.preprocessing.sequence.pad_sequences(X_train,
maxlen=128)
X_test = tf.keras.preprocessing.sequence.pad_sequences(X_test, maxlen=128)
print('X_train shape:', X_train.shape)
print('X_test shape:', X_test.shape)
print('Build model...')
model = tf.keras.Sequential()
model.add(
tf.keras.layers.Embedding(vocab_size, EMBED_SIZE, input_length=128))
model.add(tf.keras.layers.SpatialDropout1D(0.2))
model.add(
tf.keras.layers.Convolution1D(filters=nb_filter,
kernel_size=filter_length,
padding='valid',
activation='relu'))
model.add(tf.keras.layers.MaxPooling1D(pool_size=2))
model.add(tf.keras.layers.Flatten())
model.add(tf.keras.layers.Dense(HIDDEN_SIZE))
model.add(tf.keras.layers.Dropout(0.25))
model.add(tf.keras.layers.Activation(activation="relu"))
model.add(tf.keras.layers.Dense(1, activation="sigmoid"))
model.compile(optimizer='rmsprop', loss='binary_crossentropy')
model.fit(X_train,
y_train,
batch_size=BATCH_SIZE,
epochs=EPOCHS,
validation_data=(X_test, y_test))
X_pred = model.predict(X_test)
results = [result[0] for result in X_pred]
readResult(y_test, results, name="CNN", form="JSON")
return model
def cnn_improved(X_train, y_train, X_test, y_test, vocab_size):
K.clear_session()
print('X_train shape:', X_train.shape)
print('X_test shape:', X_test.shape)
print('X_train[0] before', X_train[0])
X_train = tf.keras.preprocessing.sequence.pad_sequences(X_train,
maxlen=256)
X_test = tf.keras.preprocessing.sequence.pad_sequences(X_test, maxlen=256)
print('X_train[1] after', X_train[0])
print('X_train shape:', X_train.shape)
print('X_test shape:', X_test.shape)
print('Build model...')
model = tf.keras.Sequential()
model.add(
tf.keras.layers.Embedding(vocab_size, EMBED_SIZE, input_length=256))
#model.add(tf.keras.layers.Reshape((128,256)))
print("#1:", model.output_shape)
model.add(tf.keras.layers.SpatialDropout1D(0.2))
model.add(
tf.keras.layers.Convolution1D(filters=256,
kernel_size=6,
activation='relu'))
print("#2:", model.output_shape)
model.add(tf.keras.layers.MaxPooling1D(pool_size=2))
print("#3:", model.output_shape)
model.add(
tf.keras.layers.Convolution1D(filters=256,
kernel_size=5,
activation='relu'))
print("#4:", model.output_shape)
model.add(tf.keras.layers.MaxPooling1D(pool_size=2))
print("#5:", model.output_shape)
model.add(
tf.keras.layers.Convolution1D(filters=256,
kernel_size=4,
activation='relu'))
print("#6:", model.output_shape)
model.add(tf.keras.layers.MaxPooling1D(pool_size=2))
print("#7:", model.output_shape)
model.add(
tf.keras.layers.Convolution1D(filters=256,
kernel_size=4,
activation='relu'))
print("#8:", model.output_shape)
model.add(tf.keras.layers.MaxPooling1D(pool_size=2))
print("#9:", model.output_shape)
model.add(
tf.keras.layers.Convolution1D(filters=256,
kernel_size=4,
activation='relu'))
print("#10:", model.output_shape)
model.add(tf.keras.layers.MaxPooling1D(pool_size=2))
print("#11:", model.output_shape)
model.add(tf.keras.layers.Flatten())
print("#12:", model.output_shape)
model.add(tf.keras.layers.Dense(32, activation="relu"))
print("#13:", model.output_shape)
model.add(tf.keras.layers.Dense(16, activation="relu"))
print("#14:", model.output_shape)
# model.add(tf.keras.layers.Dropout(0.25))
model.add(tf.keras.layers.Dense(1, activation="sigmoid"))
model.compile(optimizer='rmsprop', loss='binary_crossentropy')
# opt = tf.keras.optimizers.SGD(learning_rate=0.001, momentum=0.99)
#model.compile(optimizer='sgd', loss=root_mean_squared_logarithmic_error)
model.fit(X_train,
y_train,
batch_size=BATCH_SIZE,
epochs=EPOCHS,
validation_data=(X_test, y_test))
train_pred = model.predict(X_train)
train_results = [result[0] for result in train_pred]
X_pred = model.predict(X_test)
results = [result[0] for result in X_pred]
readResult(y_train, train_results, name="training_cnn+", form="JSON")
readResult(y_test, results, name="Testing_CNN+", form="JSON")
return model