hosa.optimization.hosa.HOSACNN.grid_search

HOSACNN.grid_search(n_kernels=None, n_neurons_dense_layer=None, n_units=None, n_subs_layers=None, imbalance_correction=None, **kwargs)

Runs a grid search on the remaining moldel’s parameters.

Parameters

• n_kernels (list or None) – i-th element represents the number of output filters of the convolution layer in the i-th GofLayer. Ignored in the case of optimizing an RNN.

• n_neurons_dense_layer (int or None) – Number of neurons of the penultimate dense layer (i.e., before the output layer). Ignored in the case of optimizing an CNN.

• n_units (int or None) – Dimensionality of the output space, i.e., the dimensionality of the hidden state. Ignored in the case of optimizing an CNN.

• n_subs_layers (int or None) – Ignored in the case of optimizing an CNN.

• imbalance_correction (None or bool) – Whether to apply correction to class imbalances. Only used for classification problems. Ignored for regression.

• **kwargs – Extra arguments explicitly used for regression or classification models, including the additional arguments that are used in the TensorFlow’s model fit function. See here.

Returns

tensorflow.keras.Sequential – Returns the best TensorFlow model found.