causally.scm.causal_mechanism.GaussianProcessMechanism

class causally.scm.causal_mechanism.GaussianProcessMechanism(gamma: float = 1.0)

Nonlinear causal mechanism sampled from a Gaussian process.

The nonlinear transformation is generated sampling the effect from a Gaussian process with covariance matrix defined as the kernel matrix of the parents’ observations.

Parameters:

gamma (float, default 1) – The gamma parameters fixing the variance of the kernel. Larger values of gamma determines bigger magnitude of the causal mechanisms.

Methods

predict(X)

Generate the effect given the observations of the parent nodes.