Validating Gaussian Process Models with Simulation-Based Calibration
This work addresses a specific problem for practitioners using Gaussian processes in Bayesian regression, offering a validation tool that is incremental in nature.
The paper tackles the challenge of validating Gaussian process model implementations by introducing simulation-based calibration, which successfully identified a bug in existing code and determined when hyperparameter marginalization is needed.
Gaussian process priors are a popular choice for Bayesian analysis of regression problems. However, the implementation of these models can be complex, and ensuring that the implementation is correct can be challenging. In this paper we introduce Gaussian process simulation-based calibration, a procedure for validating the implementation of Gaussian process models and demonstrate the efficacy of this procedure in identifying a bug in existing code. We also present a novel application of this procedure to identify when marginalisation of the model hyperparameters is necessary.