MultivariateNormalStructured.from_penalties()#
- classmethod MultivariateNormalStructured.from_penalties(loc, scales, penalties, tol=1e-06, precompute_masks=True, validate_args=False, allow_nan_stats=True, include_normalizing_constant=True)[source]#
Initializes the distribution directly from marginal scales and penalties (computationally expensive).
This is expensive, because it computes eigenvalue decompositions of all penalty matrices. Should only be used when performance is irrelevant.
- Return type:
Self