dnn_settings

class utilities.dnn_settings.DnnSettings(layers=(75, 60, 45, 30, 20), val_fraction=0.5, epochnum=200, learning_rate=0.001, batch_size=128, dropout=0, verbose=2)

Class to set and store some important characteristics of a Keras DNN.

Parameters:

• layers (list[int] or tuple[int]) – List or tuple of integers, that indicate the number of neurons in each internal dense layer.

• val_fraction (float) – Fraction of the training dataset used for validation.

• epochnum (float) – Number of epochs for the training.

• learning_rate (float) – Value given as learning rate to the Adam optimizer.

• batch_size (int) – Size of the batches used in the training.

• dropout (float) – Drop probability in the AlphaDropout layer.

• verbose (int) – Set how verbose the training is on shell.

Constructor method

__init__(layers=(75, 60, 45, 30, 20), val_fraction=0.5, epochnum=200, learning_rate=0.001, batch_size=128, dropout=0, verbose=2)

Constructor method

property layers

property val_fraction

property epochnum

property learning_rate

property batch_size

property dropout

property verbose