causally.scm.noise.RandomNoiseDistribution

class causally.scm.noise.RandomNoiseDistribution(standardize: bool = False)

Base abstract class for sampling from non-parametric, randomly generated, distributions.

Samples from the random distribution are generated as nonlinear transformations of samples form a standard normal. Classes inheriting from RandomNoiseDistribution must implement the _forward method, which defines the nonlinear transformation that is applied to a standard normal sample.

Methods

sample(size)

Sample random noise of the required input size.