MLP

class lightautoml.ml_algo.torch_based.nn_models.MLP(*args, **kwargs)

Bases: DenseLightModel

Realisation of ‘mlp’ model.

Parameters:

• n_in – Input dim.

• n_out – Output dim.

• hidden_size – List of hidden dims.

• drop_rate – Dropout rate for each layer separately or altogether.

• act_fun – Activation function.

• noise_std – Std of noise.

• num_init_features – If not none add fc layer before model with certain dim.

• use_bn – Use BatchNorm.

• use_noise – Use noise.

• dropout_first – Use dropout in the first layer or not.

• bn_momentum – BatchNorm momentum

• ghost_batch – If not none use GhoastNorm with ghost_batch.

• leaky_gate – Use LeakyGate or not.

• use_skip – Use another Linear model to blend them after.

• weighted_sum – Use weighted blender or half-half.

• device – Device to compute on.