depynd.markov_networks

Module contents

depynd.markov_networks.select(X, method='skeptic', criterion=None, lamb=None, verbose=False, return_lambda=False, **kwargs)

Learn the structure of Markov random field.

Parameters:

• X (array-like, shape (n_samples, n_features)) – Observations of a set of random variables.
• method ({'glasso', 'skeptic', 'gsmn', 'iamb', 'gsmple'}) – Method for structure learning.
• criterion ({'stars', 'none', None}) – Criteria for selecting regularization parameter.
• lamb (float or array-like or None) – Candidates of regularization parameter.
• verbose (bool) – If True, the objective function is plotted for each regularization parameter.
• return_lambda (bool, default False) – If True, the selected regularization parameter will be returned.
• kwargs (dict) – Optional parameters for MI estimation.

Returns:

adj – Estimated adjacency matrix of an MRF.

Return type:

array, shape (n_features, n_features)

Methods

The select() function supports the following methods:

• ‘glasso’: depynd.markov_networks._glasso()
• ‘skeptic’: depynd.markov_networks._skeptic()
• ‘gsmn’: depynd.markov_networks._gsmn()
• ‘iamb’: depynd.markov_networks._iamb()
• ‘gsmple’: depynd.markov_networks._gsmple()

Criteria

The select() function supports the following criteria:

• ‘stars’: depynd.markov_networks._stars()