Sample of FCN model

fcn_model.py

    model = Graph()
    model.add_input(name='input', input_shape=(sequence_length, features))
    #########################################################
    model.add_node(ZeroPadding1D(2), name='input_padding', input='input') # to avoid lookahead bias
    #########################################################
    model.add_node(Convolution1D(nb_filter=nb_filter, filter_length=filter_length, border_mode='valid', init=init), name='conv1', input='input_padding')
    model.add_node(ParametricSoftplus(), name='relu1', input='conv1')
    #########################################################
    model.add_node(Convolution1D(nb_filter=nb_filter, filter_length=filter_length, border_mode='same', init=init), name='conv2', input='relu1')
    model.add_node(ParametricSoftplus(), name='relu2', input='conv2')
    #########################################################
    model.add_node(Convolution1D(nb_filter=nb_filter, filter_length=filter_length, border_mode='same', init=init), name='conv3', input='relu2')
    model.add_node(ParametricSoftplus(), name='relu3', input='conv3')
    #########################################################
    model.add_node(Convolution1D(nb_filter=nb_filter, filter_length=filter_length, border_mode='same', init=init), name='conv4', input='relu3')
    model.add_node(ParametricSoftplus(), name='relu4', input='conv4')
    #########################################################
    model.add_node(Convolution1D(nb_filter=nb_filter,filter_length=filter_length, border_mode='same', init=init),
                     name='conv5',
                     inputs=['relu2', 'relu4'],
                     merge_mode='concat', concat_axis=-1)
    model.add_node(ParametricSoftplus(), name='relu5', input='conv5')
    #########################################################
    model.add_node(Convolution1D(nb_filter=nb_filter,filter_length=filter_length, border_mode='same', init=init),
                     name='conv6',
                     inputs=['relu1', 'relu5'],
                     merge_mode='concat', concat_axis=-1)
    model.add_node(ParametricSoftplus(), name='relu6', input='conv6')
    #########################################################
    model.add_node(Convolution1D(nb_filter=output_dim, filter_length=sequence_length, border_mode='same', init=init), name='conv7', input='relu6')
    model.add_node(Activation('sigmoid'), name='activation', input='conv7')
    model.add_output(name='output', input='activation')
    
    model.compile(optimizer=optimizer, loss={'output': loss})

fcn_model_working.py

    model = Graph()
    model.add_input(name='input', input_shape=(None, features))
    #########################################################
    model.add_node(ZeroPadding1D(2), name='input_padding', input='input') # to avoid lookahead bias
    #########################################################
    model.add_node(Convolution1D(nb_filter=nb_filter, filter_length=filter_length, border_mode='valid', init=init, input_shape=(sequence_length, features)), name='conv1', input='input_padding')
    model.add_node(Activation('relu'), name='relu1', input='conv1')
    #########################################################
    model.add_node(Convolution1D(nb_filter=nb_filter, filter_length=filter_length, border_mode='same', init=init), name='conv2', input='relu1')
    model.add_node(Activation('relu'), name='relu2', input='conv2')
    #########################################################
    model.add_node(Convolution1D(nb_filter=nb_filter, filter_length=filter_length, border_mode='same', init=init), name='conv3', input='relu2')
    model.add_node(Activation('relu'), name='relu3', input='conv3')
    #########################################################
    model.add_node(Convolution1D(nb_filter=nb_filter, filter_length=filter_length, border_mode='same', init=init), name='conv4', input='relu3')
    model.add_node(Activation('relu'), name='relu4', input='conv4')
    #########################################################
    model.add_node(Convolution1D(nb_filter=nb_filter,filter_length=filter_length, border_mode='same', init=init),
                     name='conv5',
                     inputs=['relu2', 'relu4'],
                     merge_mode='concat', concat_axis=-1)
    model.add_node(Activation('relu'), name='relu5', input='conv5')
    #########################################################
    model.add_node(Convolution1D(nb_filter=nb_filter,filter_length=filter_length, border_mode='same', init=init),
                     name='conv6',
                     inputs=['relu1', 'relu5'],
                     merge_mode='concat', concat_axis=-1)
    model.add_node(Activation('relu'), name='relu6', input='conv6')
    model.add_node(Convolution1D(nb_filter=output_dim, filter_length=sequence_length, border_mode='same', init=init), name='conv7', input='relu6')
    model.add_node(Activation('sigmoid'), name='activation', input='conv7')
    model.add_output(name='output', input='activation')
    
    model.compile(optimizer=optimizer, loss={'output': loss})

Model nodes shapes

input : <class 'keras.layers.core.Layer'> : (None, 500, 4) : (None, 500, 4)
input_padding : <class 'keras.layers.convolutional.ZeroPadding1D'> : (None, 500, 4) : (None, 504, 4)
conv1 : <class 'keras.layers.convolutional.Convolution1D'> : (None, 504, 4) : (None, 500, 150)
relu1 : <class 'keras.layers.advanced_activations.ParametricSoftplus'> : (None, 500, 150) : (None, 500, 150)
conv2 : <class 'keras.layers.convolutional.Convolution1D'> : (None, 500, 150) : (None, 500, 150)
relu2 : <class 'keras.layers.advanced_activations.ParametricSoftplus'> : (None, 500, 150) : (None, 500, 150)
conv3 : <class 'keras.layers.convolutional.Convolution1D'> : (None, 500, 150) : (None, 500, 150)
relu3 : <class 'keras.layers.advanced_activations.ParametricSoftplus'> : (None, 500, 150) : (None, 500, 150)
conv4 : <class 'keras.layers.convolutional.Convolution1D'> : (None, 500, 150) : (None, 500, 150)
relu4 : <class 'keras.layers.advanced_activations.ParametricSoftplus'> : (None, 500, 150) : (None, 500, 150)
conv5 : <class 'keras.layers.convolutional.Convolution1D'> : (None, 500, 300) : (None, 500, 150)
relu5 : <class 'keras.layers.advanced_activations.ParametricSoftplus'> : (None, 500, 150) : (None, 500, 150)
conv6 : <class 'keras.layers.convolutional.Convolution1D'> : (None, 500, 300) : (None, 500, 150)
relu6 : <class 'keras.layers.advanced_activations.ParametricSoftplus'> : (None, 500, 150) : (None, 500, 150)
conv7 : <class 'keras.layers.convolutional.Convolution1D'> : (None, 500, 150) : (None, 500, 3)
activation : <class 'keras.layers.core.Activation'> : (None, 500, 3) : (None, 500, 3)
output : <class 'keras.layers.core.Activation'> : (None, 500, 3) : (None, 500, 3)