setu.NRESearchSpace

class setu.NRESearchSpace(hidden_features=(25, 50, 100, 128), n_residual_blocks=(1, 5), embedding_dim=(None, 32, 64, 128), num_classes=(10, 50), learning_rate=(0.0001, 0.005))

Bases: object

Search space for NRE hyperparameter tuning.

See NLESearchSpace for the range vs categorical convention.

Parameters:

• hidden_features (tuple)

• n_residual_blocks (tuple[int, int])

• embedding_dim (tuple)

• num_classes (tuple[int, int])

• learning_rate (tuple[float, float])