DataCategorical.get_objective

DataCategorical.get_objective(**kwargs) → torch.nn.Module

Returns the loss function for approximating this feature. We use the benefit of combining the loss and activation function in one function. This avoids numerical instabilities.