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

heteroscedastic_elbo_conjugate#

gpjax.objectives.heteroscedastic_elbo_conjugate(variational_family, data)[source]#

Tight bound from Lazaro-Gredilla & Titsias (2011) for heteroscedastic Gaussian likelihoods.

Parameters:

variational_family (HVF)
data (Dataset)

Return type:

float | Float[jaxlib._jax.Array, ‘’] | Float[ndarray, ‘’]