How to implement Merge from Keras.layers

Question

I have been trying to merge the following sequential models but haven't been able to. Could somebody please point out my mistake, thank you.

The code compiles while using"merge" but give the following error "TypeError: 'module' object is not callable" However it doesn't even compile while using "Merge"

I am using keras version 2.2.0 and python 3.6

from keras.layers import merge
def linear_model_combined(optimizer='Adadelta'):    
    modela = Sequential()
    modela.add(Flatten(input_shape=(100, 34)))
    modela.add(Dense(1024))
    modela.add(Activation('relu'))
    modela.add(Dense(512))

    modelb = Sequential()
    modelb.add(Flatten(input_shape=(100, 34)))
    modelb.add(Dense(1024))
    modelb.add(Activation('relu'))
    modelb.add(Dense(512))

    model_combined = Sequential()

    model_combined.add(Merge([modela, modelb], mode='concat'))

    model_combined.add(Activation('relu'))
    model_combined.add(Dense(256))
    model_combined.add(Activation('relu'))

    model_combined.add(Dense(4))
    model_combined.add(Activation('softmax'))

    model_combined.compile(loss='categorical_crossentropy', optimizer=optimizer, metrics=['accuracy'])

    return model_combined

Answer 1

Merge cannot be used with a sequential model. In a sequential model, layers can only have one input and one output. You have to use the functional API, something like this. I assumed you use the same input layer for modela and modelb, but you could create another Input() if it is not the case and give both of them as input to the model.

def linear_model_combined(optimizer='Adadelta'):    

    # declare input
    inlayer =Input(shape=(100, 34))
    flatten = Flatten()(inlayer)

    modela = Dense(1024)(flatten)
    modela = Activation('relu')(modela)
    modela = Dense(512)(modela)

    modelb = Dense(1024)(flatten)
    modelb = Activation('relu')(modelb)
    modelb = Dense(512)(modelb)

    model_concat = concatenate([modela, modelb])


    model_concat = Activation('relu')(model_concat)
    model_concat = Dense(256)(model_concat)
    model_concat = Activation('relu')(model_concat)

    model_concat = Dense(4)(model_concat)
    model_concat = Activation('softmax')(model_concat)

    model_combined = Model(inputs=inlayer,outputs=model_concat)

    model_combined.compile(loss='categorical_crossentropy', optimizer=optimizer, metrics=['accuracy'])

    return model_combined

Answer 2

The keras.layers.merge layer is deprecated. Use keras.layers.Concatenate(axis=-1) instead as mentioned here: https://keras.io/layers/merge/#concatenate

Answer 3

To be honest, I was struggling on this issue for a long time...

Luckily I found the panacea expected finally. For anyone who would like to make the minimal changes on their original codes with Sequential, here comes the solution:

def linear_model_combined(optimizer='Adadelta'): 
    from keras.models import Model, Sequential
    from keras.layers.core import Dense, Flatten, Activation, Dropout
    from keras.layers import add

    modela = Sequential()
    modela.add(Flatten(input_shape=(100, 34)))
    modela.add(Dense(1024))
    modela.add(Activation('relu'))
    modela.add(Dense(512))

    modelb = Sequential()
    modelb.add(Flatten(input_shape=(100, 34)))
    modelb.add(Dense(1024))
    modelb.add(Activation('relu'))
    modelb.add(Dense(512))

    merged_output = add([modela.output, modelb.output])   

    model_combined = Sequential()
    model_combined.add(Activation('relu'))
    model_combined.add(Dense(256))
    model_combined.add(Activation('relu'))
    model_combined.add(Dense(4))
    model_combined.add(Activation('softmax'))

    final_model = Model([modela.input, modelb.input], model_combined(merged_output))

    final_model.compile(loss='categorical_crossentropy', optimizer=optimizer, metrics=['accuracy'])

    return final_model

For more information, refer to https://github.com/keras-team/keras/issues/3921#issuecomment-335457553 for farizrahman4u's comment. ;)