Infinite Shift-invariant Grouped Multi-task Learning for Gaussian Processes

TL;DR

This paper introduces an innovative Bayesian nonparametric multi-task learning model for phase-shifted periodic time series, utilizing Gaussian processes and Dirichlet Process priors to improve modeling and inference in complex, sparse, and non-synchronous data scenarios.

Contribution

It develops a novel infinite mixture model with an EM algorithm and Variational Bayesian inference for phase-shifted Gaussian process time series, enabling automatic model selection and improved performance.

Findings

Effective in regression, classification, and class discovery tasks.

Performs well on synthetic and astrophysics real-world data.

Handles sparse and non-synchronous time series effectively.

Abstract

Multi-task learning leverages shared information among data sets to improve the learning performance of individual tasks. The paper applies this framework for data where each task is a phase-shifted periodic time series. In particular, we develop a novel Bayesian nonparametric model capturing a mixture of Gaussian processes where each task is a sum of a group-specific function and a component capturing individual variation, in addition to each task being phase shifted. We develop an efficient \textsc{em} algorithm to learn the parameters of the model. As a special case we obtain the Gaussian mixture model and \textsc{em} algorithm for phased-shifted periodic time series. Furthermore, we extend the proposed model by using a Dirichlet Process prior and thereby leading to an infinite mixture model that is capable of doing automatic model selection. A Variational Bayesian approach is developed for inference in this model. Experiments in regression, classification and class discovery demonstrate the performance of the proposed models using both synthetic data and real-world time series data from astrophysics. Our methods are particularly useful when the time series are sparsely and non-synchronously sampled.