nn.py 文件源码

def cnn_combine_train(X_train_list,y_train,vocab_size):
    N=len(X_train_list)

    X_train_list = [sequence.pad_sequences(x_train, maxlen=MAX_LEN) for x_train in X_train_list]

    input_list=[]
    out_list=[]
    for i in range(N):
        input,out=get_embedding_input_output('f%d' %i,vocab_size)
        input_list.append(input)
        out_list.append(out)

    x = merge(out_list,mode='concat')

    x = Dropout(0.25)(x)

    # we add a Convolution1D, which will learn nb_filter
    # word group filters of size filter_length:
    x = Convolution1D(nb_filter=nb_filter,
                            filter_length=filter_length,
                            border_mode='valid',
                            activation='relu',
                            subsample_length=1)(x)

    # we use standard max pooling (halving the output of the previous layer):
    x = MaxPooling1D(pool_length=2)(x)

    # We flatten the output of the conv layer,
    # so that we can add a vanilla dense layer:
    x = Flatten()(x)

    # We add a vanilla hidden layer:
    x = Dense(HIDDEN_SIZE)(x)
    x = Dropout(0.25)(x)
    x = Activation('relu')(x)

    # We project onto a single unit output layer, and squash it with a sigmoid:
    x = Dense(1)(x)
    x = Activation('sigmoid')(x)

    model = Model(input=input_list, output=x)

    model.compile(loss='binary_crossentropy', optimizer='rmsprop')
    model.fit(X_train_list, y_train, batch_size=BATCH_SIZE, nb_epoch=EPOCHS)

    return model