Fast Gaussian Processes under Monotonicity Constraints

TL;DR

This paper introduces a novel, efficient framework for Gaussian process models with monotonicity constraints, improving computational speed and applicability in high-dimensional scientific problems.

Contribution

It proposes a virtual point-based approach using RLRTO and NUTS sampling, enhancing efficiency and scalability of constrained Gaussian process models.

Findings

Comparable predictive performance across methods

Significant computational efficiency improvements

Effective application to differential equation systems

Abstract

Gaussian processes (GPs) are widely used as surrogate models for complicated functions in scientific and engineering applications. In many cases, prior knowledge about the function to be approximated, such as monotonicity, is available and can be leveraged to improve model fidelity. Incorporating such constraints into GP models enhances predictive accuracy and reduces uncertainty, but remains a computationally challenging task for high-dimensional problems. In this work, we present a novel virtual point-based framework for building constrained GP models under monotonicity constraints, based on regularized linear randomize-then-optimize (RLRTO), which enables efficient sampling from a constrained posterior distribution by means of solving randomized optimization problems. We also enhance two existing virtual point-based approaches by replacing Gibbs sampling with the No U-Turn Sampler (NUTS) for improved efficiency. A Python implementation of these methods is provided and can be easily applied to a wide range of problems. This implementation is then used to validate the approaches on approximating a range of synthetic functions, demonstrating comparable predictive performance between all considered methods and significant improvements in computational efficiency with the two NUTS methods and especially with the RLRTO method. The framework is further applied to construct surrogate models for systems of differential equations.