String Gaussian Process Kernels
This work addresses the challenge of flexible nonstationary modeling in Gaussian processes for researchers and practitioners in machine learning, representing an incremental improvement with a novel kernel class.
The authors tackled the problem of modeling nonstationary data with multiple local patterns by introducing string Gaussian processes and their kernels, demonstrating that the model outperforms competing approaches on real-life datasets with nonstationary features.
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.