medil.models.GaussianMCM

class medil.models.GaussianMCM(**kwargs)[source]

Bases: _MedilCausalModel

Linear Gaussian MeDIL causal model.

Learns a bipartite latent→measurement causal structure and estimates linear Gaussian parameters (edge weights, error means, error variances) by constraint-based structure learning and least-squares optimization of the covariance matrix.

Parameters:

biadj (ndarray of shape (num_latent, num_meas), optional) – Boolean biadjacency matrix. If empty (default), estimated from data during fit().
udg (ndarray of shape (num_meas, num_meas), optional) – Boolean undirected dependence graph over observed variables. If empty (default), estimated from data during fit().
rng (numpy.random.Generator, optional) – Random number generator used during sample().

biadj

Boolean biadjacency matrix (set after fit() or at init).

Type:: ndarray of shape (num_latent, num_meas)

parameters

Learned parameters with attributes biadj_weights (shape (num_latent, num_meas)), error_means (shape (num_meas,)), and error_variances (shape (num_meas,)).

Type:: object

Methods

`__init__`(**kwargs)
`fit`(dataset)	Fit a GaussianMCM to a dataset.
`sample`(sample_size[, include_latent])	Sample observations from a GaussianMCM.