从高斯过程到高斯过程混合模型:研究与展望

From Gaussian Processes to the Mixture of Gaussian Processes: A Survey

高斯过程(GP)模型是核学习方法与贝叶斯推理相结合的典范,现已成为机器学习领域的一个研究热点。作为对GP模型的拓展,高斯过程混合(MGP)模型具有更强大的学习能力和适应性。然而,目前关于GP和MGP模型的研究较为零散,尚缺少系统的分析与总结。本文首先对于GP模型的基本原理及其研究进展进行了深入地分析和讨论;然后将GP模型拓展至MGP模型,从多方面对MGP模型的研究现状和进展进行了深入地分析和讨论,并指出未来值得探索的研究方向和应用问题。

Gaussian process （GP） model is a paradigmatic machine learning model that combines the advantages of both kernel learning method and Bayesian inference mechanism, and thus has become a very popular area in machine learning in recent years. As an extension of the GP model, the Mixture of Gaussian Processes （MGP） fits datasets more effectively and thus it has a better ability of learning and generalization. However, there are only some isolated literatures and reports about the GP and MGP models and no systematic summary on these models. In this paper, we begin to review the GP model and its basic principles and developments on various aspects. We then discuss how to extend the GP model to the MGP model and further review the status and developments of the MGP models, and finally point out some prospective research direc- tions and interesting applications of the MGP model.