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

NonNegativeReal#

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

Bases: AbstractUnwrappable

Non-negative parameter (semantically allows zero, e.g. jitter, noise floor).

Uses the same softplus bijection as PositiveReal. The distinction is semantic: NonNegativeReal signals that zero is a meaningful boundary.

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