lstm_with_dropout

lstm_with_dropout.py

def lstm_net_with_dropout(input_shape, output_dim, dropout=0.15):
    inputs = keras.Input(shape=input_shape)
    x = keras.layers.LSTM(30, recurrent_dropout=dropout, return_sequences=True)(inputs, training=True)
    x = keras.layers.Dropout(dropout)(x, training=True)
    x = keras.layers.LSTM(30)(x, training=True)
    x = keras.layers.Dropout(dropout)(x, training=True)
    outputs = keras.layers.Dense(output_dim)(x)
    model = keras.Model(inputs, outputs)
    model.compile(loss='mean_squared_error', optimizer='adam', metrics=['mae', 'mse']) 
    return model