This module contains helper functions and classes that allow you to prototype quickly or that can be used for writing tests.
.MLPModule
is a simple PyTorch ~torch.nn.Module
that implements a multi-layer perceptron. It allows to indicate the number of input, hidden, and output units, as well as the non-linearity and use of dropout. You can use this module directly in conjunction with .NeuralNet
.
Additionally, the functions ~skorch.toy.make_classifier
, ~skorch.toy.make_binary_classifier
, and ~skorch.toy.make_regressor
can be used to return a .MLPModule
with the defaults adjusted for use in multi-class classification, binary classification, and regression, respectively.