自学教程：Python layers.ELU属性代码示例

# 需要导入模块: from keras import layers [as 别名]# 或者: from keras.layers import ELU [as 别名]def get_model(time_len=1):  ch, row, col = 3, 160, 320  # camera format  model = Sequential()  model.add(Lambda(lambda x: x/127.5 - 1.,            input_shape=(ch, row, col),            output_shape=(ch, row, col)))  model.add(Convolution2D(16, 8, 8, subsample=(4, 4), border_mode="same"))  model.add(ELU())  model.add(Convolution2D(32, 5, 5, subsample=(2, 2), border_mode="same"))  model.add(ELU())  model.add(Convolution2D(64, 5, 5, subsample=(2, 2), border_mode="same"))  model.add(Flatten())  model.add(Dropout(.2))  model.add(ELU())  model.add(Dense(512))  model.add(Dropout(.5))  model.add(ELU())  model.add(Dense(1))  model.compile(optimizer="adam", loss="mse")  return model 

# 需要导入模块: from keras import layers [as 别名]# 或者: from keras.layers import ELU [as 别名]def Encoder(hidden_size, activation=None, return_sequences=True, bidirectional=False, use_gru=True):    if activation is None:        activation = ELU()    if use_gru:        def _encoder(x):            if bidirectional:                branch_1 = GRU(int(hidden_size/2), activation='linear',                               return_sequences=return_sequences, go_backwards=False)(x)                branch_2 = GRU(int(hidden_size/2), activation='linear',                               return_sequences=return_sequences, go_backwards=True)(x)                x = concatenate([branch_1, branch_2])                x = activation(x)                return x            else:                x = GRU(hidden_size, activation='linear',                        return_sequences=return_sequences)(x)                x = activation(x)                return x    else:        def _encoder(x):            if bidirectional:                branch_1 = LSTM(int(hidden_size/2), activation='linear',                                return_sequences=return_sequences, go_backwards=False)(x)                branch_2 = LSTM(int(hidden_size/2), activation='linear',                                return_sequences=return_sequences, go_backwards=True)(x)                x = concatenate([branch_1, branch_2])                x = activation(x)                return x            else:                x = LSTM(hidden_size, activation='linear',                         return_sequences=return_sequences)(x)                x = activation(x)                return x    return _encoder 

# 需要导入模块: from keras import layers [as 别名]# 或者: from keras.layers import ELU [as 别名]def Decoder(hidden_size, activation=None, return_sequences=True, bidirectional=False, use_gru=True):    if activation is None:        activation = ELU()    if use_gru:        def _decoder(x):            if bidirectional:                x = Bidirectional(                    GRU(int(hidden_size/2), activation='linear', return_sequences=return_sequences))(x)                x = activation(x)                return x            else:                x = GRU(hidden_size, activation='linear',                        return_sequences=return_sequences)(x)                x = activation(x)                return x    else:        def _decoder(x):            if bidirectional:                x = Bidirectional(                    LSTM(int(hidden_size/2), activation='linear', return_sequences=return_sequences))(x)                x = activation(x)                return x            else:                x = LSTM(hidden_size, activation='linear',                         return_sequences=return_sequences)(x)                x = activation(x)                return x    return _decoder 

# 需要导入模块: from keras import layers [as 别名]# 或者: from keras.layers import ELU [as 别名]def test_elu(self):        keras_model = Sequential()        keras_model.add(ELU(input_shape=(3, 32, 32), name='elu'))        keras_model.compile(loss=keras.losses.categorical_crossentropy,                            optimizer=keras.optimizers.SGD())        pytorch_model = ELUNet()        self.transfer(keras_model, pytorch_model)        self.assertEqualPrediction(keras_model, pytorch_model, self.test_data)    # Tests activation function with learned parameters 

# 需要导入模块: from keras import layers [as 别名]# 或者: from keras.layers import ELU [as 别名]def convresblock(x, nfeats=8, ksize=3, nskipped=2, elu=True):    """The proposed residual block from [4].    Running with elu=True will use ELU nonlinearity and running with    elu=False will use BatchNorm + RELU nonlinearity.  While ELU's are fast    due to the fact they do not suffer from BatchNorm overhead, they may    overfit because they do not offer the stochastic element of the batch    formation process of BatchNorm, which acts as a good regularizer.    # Arguments        x: 4D tensor, the tensor to feed through the block        nfeats: Integer, number of feature maps for conv layers.        ksize: Integer, width and height of conv kernels in first convolution.        nskipped: Integer, number of conv layers for the residual function.        elu: Boolean, whether to use ELU or BN+RELU.    # Input shape        4D tensor with shape:        `(batch, channels, rows, cols)`    # Output shape        4D tensor with shape:        `(batch, filters, rows, cols)`    """    y0 = Conv2D(nfeats, ksize, padding='same')(x)    y = y0    for i in range(nskipped):        if elu:            y = ELU()(y)        else:            y = BatchNormalization(axis=1)(y)            y = Activation('relu')(y)        y = Conv2D(nfeats, 1, padding='same')(y)    return layers.add([y0, y]) 

# 需要导入模块: from keras import layers [as 别名]# 或者: from keras.layers import ELU [as 别名]def test_elu():    for alpha in [0., .5, -1.]:        layer_test(layers.ELU, kwargs={'alpha': alpha},                   input_shape=(2, 3, 4)) 

# 需要导入模块: from keras import layers [as 别名]# 或者: from keras.layers import ELU [as 别名]def create_model(input_shape, hidden_layers=[1024, 512, 256], input_dropout=0.1, hidden_dropout=0.5):    '''Define a simple multilayer perceptron.    Args:        input_shape (tuple): input shape to the model. For this model, should be of shape (dim,)        input_dropout (float): fraction of input features to drop out during training        hidden_layers (tuple): a tuple/list with number of hidden units in each hidden layer    Returns:        keras.models.Sequential : a model to train    '''    model = Sequential()    # dropout the input to prevent overfitting to any one feature    # (a similar concept to randomization in random forests,    #   but we choose less severe feature sampling  )    model.add(Dropout(input_dropout, input_shape=input_shape))    # set up hidden layers    for n_hidden_units in hidden_layers:        # the layer...activation will come later        model.add(Dense(n_hidden_units))        # dropout to prevent overfitting        model.add(Dropout(hidden_dropout))        # batchnormalization helps training        model.add(BatchNormalization())        # ...the activation!        model.add(ELU())    # the output layer    model.add(Dense(1, activation='sigmoid'))    # we'll optimize with plain old sgd    model.compile(loss='binary_crossentropy',                  optimizer='sgd', metrics=['accuracy'])    return model 

# 需要导入模块: from keras import layers [as 别名]# 或者: from keras.layers import ELU [as 别名]def ResidualBlock1D_helper(layers, kernel_size, filters, final_stride=1):    def f(_input):        basic = _input        for ln in range(layers):            #basic = BatchNormalization()( basic ) # triggers known keras bug w/ TimeDistributed: https://github.com/fchollet/keras/issues/5221            basic = ELU()(basic)              basic = Conv1D(filters, kernel_size, kernel_initializer='he_normal',                           kernel_regularizer=l2(1.e-4), padding='same')(basic)        # note that this strides without averaging        return AveragePooling1D(pool_size=1, strides=final_stride)(Add()([_input, basic]))    return f 

# 需要导入模块: from keras import layers [as 别名]# 或者: from keras.layers import ELU [as 别名]def build_model(output_size):    channel_axis = 3    freq_axis = 1    padding = 37    input_shape = (img_height, img_width, channels)    print('Building model...')    model = Sequential()    #model.add(ZeroPadding2D(padding=(0, padding), data_format='channels_last', input_shape=input_shape))    #model.add(BatchNormalization(axis=freq_axis, name='bn_0_freq'))    #model.add(Conv2D(64, (3, 3), padding='same', name='conv1'))    #model.add(BatchNormalization(axis=channel_axis, name='bn1'))    #model.add(ELU())    #model.add(MaxPooling2D(pool_size=(2, 2), strides=(2, 2), name='pool1'))    #model.add(Dropout(0.1, name='dropout1'))    #model.add(Conv2D(128, (3, 3), padding='same', name='conv2'))    #model.add(BatchNormalization(axis=channel_axis, name='bn2'))    #model.add(ELU())    #model.add(MaxPooling2D(pool_size=(3, 3), strides=(3, 3), name='pool2'))    #model.add(Dropout(0.1, name='dropout2'))    #model.add(Conv2D(128, (3, 3), padding='same', name='conv3'))    #model.add(BatchNormalization(axis=channel_axis, name='bn3'))    #model.add(ELU())    #model.add(MaxPooling2D(pool_size=(4, 4), strides=(4, 4), name='pool3'))    #model.add(Dropout(0.1, name='dropout3'))    #model.add(Conv2D(128, (3, 3), padding='same', name='conv4'))    #model.add(BatchNormalization(axis=channel_axis, name='bn4'))    #model.add(ELU())    #model.add(MaxPooling2D(pool_size=(4, 4), strides=(4, 4), name='pool4'))    #model.add(Dropout(0.1, name='dropout4'))    #model.add(Reshape(target_shape=(15, 128)))    #model.add(GRU(32, return_sequences=True, name='gru1'))    #model.add(GRU(32, return_sequences=False, name='gru2'))    #model.add(Dropout(0.3, name='dropout_final'))    model.add(Reshape(target_shape=(img_height * img_width,), input_shape=input_shape))    model.add(Dense(output_size, activation='softmax', name='output', input_shape=input_shape))    return model 

# 需要导入模块: from keras import layers [as 别名]# 或者: from keras.layers import ELU [as 别名]def build_model(output_size):    channel_axis = 3    freq_axis = 1    padding = 37    input_shape = (img_height, img_width, channels)    print('Building model...')    model = Sequential()    model.add(ZeroPadding2D(padding=(0, padding), data_format='channels_last', input_shape=input_shape))    model.add(BatchNormalization(axis=freq_axis, name='bn_0_freq'))    model.add(Conv2D(64, (3, 3), padding='same', name='conv1'))    model.add(BatchNormalization(axis=channel_axis, name='bn1'))    model.add(ELU())    model.add(MaxPooling2D(pool_size=(2, 2), strides=(2, 2), name='pool1'))    model.add(Dropout(0.1, name='dropout1'))    model.add(Conv2D(128, (3, 3), padding='same', name='conv2'))    model.add(BatchNormalization(axis=channel_axis, name='bn2'))    model.add(ELU())    model.add(MaxPooling2D(pool_size=(3, 3), strides=(3, 3), name='pool2'))    model.add(Dropout(0.1, name='dropout2'))    model.add(Conv2D(128, (3, 3), padding='same', name='conv3'))    model.add(BatchNormalization(axis=channel_axis, name='bn3'))    model.add(ELU())    model.add(MaxPooling2D(pool_size=(4, 4), strides=(4, 4), name='pool3'))    model.add(Dropout(0.1, name='dropout3'))    model.add(Conv2D(128, (3, 3), padding='same', name='conv4'))    model.add(BatchNormalization(axis=channel_axis, name='bn4'))    model.add(ELU())    model.add(MaxPooling2D(pool_size=(4, 4), strides=(4, 4), name='pool4'))    model.add(Dropout(0.1, name='dropout4'))    model.add(Reshape(target_shape=(15, 128)))    model.add(GRU(32, return_sequences=True, name='gru1'))    model.add(GRU(32, return_sequences=False, name='gru2'))    model.add(Dropout(0.3, name='dropout_final'))    model.add(Dense(output_size, activation='softmax', name='output'))    return model 

# 需要导入模块: from keras import layers [as 别名]# 或者: from keras.layers import ELU [as 别名]def buildModel(cameraFormat=(3, 480, 640)):  """  Build and return a CNN; details in the comments.  The intent is a scaled down version of the model from "End to End Learning  for Self-Driving Cars": https://arxiv.org/abs/1604.07316.  Args:    cameraFormat: (3-tuple) Ints to specify the input dimensions (color        channels, rows, columns).  Returns:    A compiled Keras model.  """  print "Building model..."  ch, row, col = cameraFormat  model = Sequential()  # Use a lambda layer to normalize the input data  model.add(Lambda(      lambda x: x/127.5 - 1.,      input_shape=(ch, row, col),      output_shape=(ch, row, col))  )  # Several convolutional layers, each followed by ELU activation  # 8x8 convolution (kernel) with 4x4 stride over 16 output filters  model.add(Convolution2D(16, 8, 8, subsample=(4, 4), border_mode="same"))  model.add(ELU())  # 5x5 convolution (kernel) with 2x2 stride over 32 output filters  model.add(Convolution2D(32, 5, 5, subsample=(2, 2), border_mode="same"))  model.add(ELU())  # 5x5 convolution (kernel) with 2x2 stride over 64 output filters  model.add(Convolution2D(64, 5, 5, subsample=(2, 2), border_mode="same"))  # Flatten the input to the next layer  model.add(Flatten())  # Apply dropout to reduce overfitting  model.add(Dropout(.2))  model.add(ELU())  # Fully connected layer  model.add(Dense(512))  # More dropout  model.add(Dropout(.5))  model.add(ELU())  # Fully connected layer with one output dimension (representing the speed).  model.add(Dense(1))  # Adam optimizer is a standard, efficient SGD optimization method  # Loss function is mean squared error, standard for regression problems  model.compile(optimizer="adam", loss="mse")  return model