TrainData_deepFlavour.py 文件源码

def base_model(input_shapes):
        from keras.layers import Input
        from keras.layers.core import Masking
        x_global  = Input(shape=input_shapes[0])
        x_map = Input(shape=input_shapes[1])
        x_ptreco  = Input(shape=input_shapes[2])

        x =   Convolution2D(64, (8,8)  , border_mode='same', activation='relu',kernel_initializer='lecun_uniform')(x_map)
        x = MaxPooling2D(pool_size=(2, 2))(x)
        x =   Convolution2D(64, (4,4) , border_mode='same', activation='relu',kernel_initializer='lecun_uniform')(x)
        x = MaxPooling2D(pool_size=(2, 2))(x)
        x =   Convolution2D(64, (4,4)  , border_mode='same', activation='relu',kernel_initializer='lecun_uniform')(x)
        x = MaxPooling2D(pool_size=(2, 2))(x)
        x = Flatten()(x)
        x = merge( [x, x_global] , mode='concat')
        # linear activation for regression and softmax for classification
        x = Dense(128, activation='relu',kernel_initializer='lecun_uniform')(x)
        x = merge([x, x_ptreco], mode='concat')
        return [x_global, x_map, x_ptreco], x