mlair.model_modules.flatten

Module Contents

Functions

get_activation(input_to_activate: tensorflow.keras.layers, activation: Union[Callable, str], **kwargs)	Apply activation on a given input layer.
flatten_tail(input_x: tensorflow.keras.layers, inner_neurons: int, activation: Union[Callable, str], output_neurons: int, output_activation: Union[Callable, str], reduction_filter: int = None, name: str = None, bound_weight: bool = False, dropout_rate: float = None, kernel_regularizer: tensorflow.keras.regularizers = None)	Flatten output of convolutional layers.

Attributes

__author__
__date__

mlair.model_modules.flatten.__author__ = Felix Kleinert, Lukas Leufen

mlair.model_modules.flatten.__date__ = 2019-12-02

mlair.model_modules.flatten.get_activation(input_to_activate: tensorflow.keras.layers, activation: Union[Callable, str], **kwargs)

Apply activation on a given input layer.

This helper function is able to handle advanced keras activations as well as strings for standard activations.

Parameters

• input_to_activate – keras layer to apply activation on

• activation – activation to apply on `input_to_activate’. Can be a standard keras strings or activation layers

• kwargs – keyword arguments used inside activation layer

Returns

activation

input_x = ... # your input data
x_in = keras.layer(<without activation>)(input_x)

# get activation via string
x_act_string = get_activation(x_in, 'relu')
# or get activation via layer callable
x_act_layer = get_activation(x_in, keras.layers.advanced_activations.ELU)

mlair.model_modules.flatten.flatten_tail(input_x: tensorflow.keras.layers, inner_neurons: int, activation: Union[Callable, str], output_neurons: int, output_activation: Union[Callable, str], reduction_filter: int = None, name: str = None, bound_weight: bool = False, dropout_rate: float = None, kernel_regularizer: tensorflow.keras.regularizers = None)

Flatten output of convolutional layers.

Parameters

• input_x – Multidimensional keras layer (ConvLayer)

• output_neurons – Number of neurons in the last layer (must fit the shape of labels)

• output_activation – final activation function

• name – Name of the flatten tail.

• bound_weight – Use `tanh’ as inner activation if set to True, otherwise `activation’

• dropout_rate – Dropout rate to be applied between trainable layers

• activation – activation to after conv and dense layers

• reduction_filter – number of filters used for information compression on `input_x’ before flatten()

• inner_neurons – Number of neurons in inner dense layer

• kernel_regularizer – regularizer to apply on conv and dense layers

Returns

flatten branch with size n=output_neurons

input_x = ... # your input data
conv_out = Conv2D(*args)(input_x) # your convolution stack
out = flatten_tail(conv_out, inner_neurons=64, activation=keras.layers.advanced_activations.ELU,
                   output_neurons=4
                   output_activation='linear', reduction_filter=64,
                   name='Main', bound_weight=False, dropout_rate=.3,
                   kernel_regularizer=keras.regularizers.l2()
                   )
model = keras.Model(inputs=input_x, outputs=[out])