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

stabilised_cholesky#

gpjax.linalg.stabilised_cholesky(matrix, jitter)[source]#

Lower Cholesky factor of matrix + jitter * I.

The single stabilise-and-factor entry point for GP conditioning: the jitter policy is applied here and nowhere else. Structure-aware dispatch over lineax operators arrives with the linalg deepening; the interface will not change.

Parameters:

matrix (Array)
jitter (float | Array)

Return type:

Array