String Gaussian Process Kernels

TL;DR

This paper introduces string Gaussian process kernels, a new nonstationary kernel class derived from string GPs, enabling modeling of complex local patterns with global regularity, and demonstrates superior performance on real datasets.

Contribution

The paper presents a novel class of nonstationary kernels based on string GPs, allowing flexible local pattern modeling with global regularity constraints.

Findings

Outperforms existing methods on real nonstationary datasets

Effectively models multiple local patterns in data

Demonstrates advantages with synthetic and real data

Abstract

We introduce a new class of nonstationary kernels, which we derive as covariance functions of a novel family of stochastic processes we refer to as string Gaussian processes (string GPs). We construct string GPs to allow for multiple types of local patterns in the data, while ensuring a mild global regularity condition. In this paper, we illustrate the efficacy of the approach using synthetic data and demonstrate that the model outperforms competing approaches on well studied, real-life datasets that exhibit nonstationary features.