phygnn.layers.handlers.Layers

class Layers(n_features, n_labels=1, hidden_layers=None, input_layer=None, output_layer=None)

Bases: HiddenLayers

Class to handle TensorFlow layers

Parameters:

• n_features (int) – Number of features (inputs) to train the model on

• n_labels (int, optional) – Number of labels (outputs) to the model, by default 1

• hidden_layers (list | None, optional) – List of dictionaries of key word arguments for each hidden layer in the NN. Dense linear layers can be input with their activations or separately for more explicit control over the layer ordering. For example, this is a valid input for hidden_layers that will yield 8 hidden layers (10 layers including input+output):

  [{‘units’: 64, ‘activation’: ‘relu’, ‘dropout’: 0.01},

  {‘units’: 64}, {‘batch_normalization’: {‘axis’: -1}}, {‘activation’: ‘relu’}, {‘dropout’: 0.01}, {‘class’: ‘Flatten’}, ]

  by default None which will lead to a single linear layer

• input_layer (None | bool | dict) – Input layer. specification. Can be a dictionary similar to hidden_layers specifying a dense / conv / lstm layer. Will default to a keras InputLayer with input shape = n_features. Can be False if the input layer will be included in the hidden_layers input.

• output_layer (None | bool | list | dict) – Output layer specification. Can be a list/dict similar to hidden_layers input specifying a dense layer with activation. For example, for a classfication problem with a single output, output_layer should be [{‘units’: 1}, {‘activation’: ‘sigmoid’}]. This defaults to a single dense layer with no activation (best for regression problems). Can be False if the output layer will be included in the hidden_layers input.

Methods

add_layer(layer_kwargs)	Add a hidden layer to the DNN.
add_layer_by_class(class_name, **kwargs)	Add a new layer by the class name, either from phygnn.layers.custom_layers or tf.keras.layers
add_skip_layer(name)	Add a skip layer, looking for a prior skip connection start point if already in the layer list.
compile(model, n_features[, n_labels, ...])	Build all layers needed for model
parse_repeats(hidden_layers)	Parse repeat layers.

Attributes

bias_weights	Get a list of the NN bias weights (tensors)
hidden_layer_kwargs	List of dictionaries of key word arguments for each hidden layer in the NN.
input_layer_kwargs	Dictionary of key word arguments for the input layer.
kernel_weights	Get a list of the NN kernel weights (tensors)
layers	TensorFlow keras layers
output_layer_kwargs	Dictionary of key word arguments for the output layer.
skip_layers	Get a dictionary of unique SkipConnection objects in the layers list keyed by SkipConnection name.
weights	Get a list of layer weights for gradient calculations.

property input_layer_kwargs

Dictionary of key word arguments for the input layer. This is a copy of the input_layer input arg that can be used to reconstruct the network.

Returns:

list

add_layer(layer_kwargs)

Add a hidden layer to the DNN.

Parameters:

layer_kwargs (dict) – Dictionary of key word arguments for list layer. For example, any of the following are valid inputs:

{‘units’: 64, ‘activation’: ‘relu’, ‘dropout’: 0.05} {‘units’: 64, ‘name’: ‘relu1’} {‘activation’: ‘relu’} {‘batch_normalization’: {‘axis’: -1}} {‘dropout’: 0.1}

add_layer_by_class(class_name, **kwargs)

Add a new layer by the class name, either from phygnn.layers.custom_layers or tf.keras.layers

Parameters:

• class_name (str) – Class name from phygnn.layers.custom_layers or tf.keras.layers

• kwargs (dict) – Key word arguments to initialize the class.

add_skip_layer(name)

Add a skip layer, looking for a prior skip connection start point if already in the layer list.

Parameters:

name (str) – Unique string identifier of the skip connection. The skip endpoint should have the same name.

property bias_weights

Get a list of the NN bias weights (tensors)

(can be used for bias regularization).

Does not include input layer or dropout layers. Does include the output layer.

Returns:

list

property hidden_layer_kwargs

List of dictionaries of key word arguments for each hidden layer in the NN. This is a copy of the hidden_layers input arg that can be used to reconstruct the network.

Returns:

list

property kernel_weights

Get a list of the NN kernel weights (tensors)

(can be used for kernel regularization).

Does not include input layer or dropout layers. Does include the output layer.

Returns:

list

property layers

TensorFlow keras layers

Returns:

list

property output_layer_kwargs

Dictionary of key word arguments for the output layer. This is a copy of the output_layer input arg that can be used to reconstruct the network.

Returns:

list

static parse_repeats(hidden_layers)

Parse repeat layers. Must have “repeat” and “n” to repeat one or more layers.

Parameters:

hidden_layers (list) – Hidden layer kwargs including possibly entries with {‘n’: 2, ‘repeat’: [{…}, {…}]} that will duplicate the list sub entry n times.

Returns:

hidden_layers (list) – Hidden layer kwargs exploded for ‘repeat’ entries.

property skip_layers

Get a dictionary of unique SkipConnection objects in the layers list keyed by SkipConnection name.

Returns:

list

property weights

Get a list of layer weights for gradient calculations.

Returns:

list

classmethod compile(model, n_features, n_labels=1, hidden_layers=None, input_layer=None, output_layer=None)

Build all layers needed for model

Parameters:

• model (tensorflow.keras.Sequential) – Model to add layers too

• n_features (int) – Number of features (inputs) to train the model on

• n_labels (int, optional) – Number of labels (outputs) to the model, by default 1

• hidden_layers (list | None, optional) – List of dictionaries of key word arguments for each hidden layer in the NN. Dense linear layers can be input with their activations or separately for more explicit control over the layer ordering. For example, this is a valid input for hidden_layers that will yield 7 hidden layers (9 layers total):

  [{‘units’: 64, ‘activation’: ‘relu’, ‘dropout’: 0.01},

  {‘units’: 64}, {‘batch_normalization’: {‘axis’: -1}}, {‘activation’: ‘relu’}, {‘dropout’: 0.01}]

  by default None which will lead to a single linear layer

• input_layer (None | bool | InputLayer) – Keras input layer. Will default to an InputLayer with input shape = n_features. Can be False if the input layer will be included in the hidden_layers input.

• output_layer (None | bool | list | dict) – Output layer specification. Can be a list/dict similar to hidden_layers input specifying a dense layer with activation. For example, for a classfication problem with a single output, output_layer should be [{‘units’: 1}, {‘activation’: ‘sigmoid’}] This defaults to a single dense layer with no activation (best for regression problems). Can be False if the output layer will be included in the hidden_layers input.

Returns:

model (tensorflow.keras) – Model with layers added