Smoke render: optimisation and MCMC are shrunk for CI speed, so figures and diagnostics are not publication-fidelity. The production docs are built with full-length inference.

GPJax

LowerTriangular#

class gpjax.parameters.LowerTriangular(value)[source]#

Bases: AbstractUnwrappable

Lower-triangular matrix parameter with positive diagonal (Cholesky factor).

Stored as a flat vector; unwrap() fills a lower-triangular matrix with softplus applied to the diagonal entries.

unwrap()[source]#: Returns the unwrapped pytree, assuming no wrapped subnodes exist.