Standard Monte Carlo algorithm for hierarchical Gaussian‑process models

Establish a standard Monte Carlo algorithm for general hierarchical Gaussian‑process models that can be broadly applied beyond restricted special cases.

Background

The paper surveys existing inference approaches for hierarchical Gaussian‑process models and notes that variational methods can introduce bias, while available Monte Carlo methods either rely on restrictive assumptions (e.g., Gaussian likelihoods) or lack demonstrated efficiency and scalability.

Within this context, the authors explicitly point out that, despite extensive interest and partial solutions, a generally accepted standard Monte Carlo algorithm covering the broad class of hierarchical GP models is still missing.