Deep Structured Mixtures of Gaussian Processes

Martin Trapp, Robert Peharz, Franz Pernkopf, Carl E. Rasmussen

    Onderzoeksoutput: Hoofdstuk in Boek/Rapport/CongresprocedureConferentiebijdrageAcademicpeer review


    Gaussian Processes (GPs) are powerful non-parametric Bayesian regression models that allow exact posterior inference, but exhibit high computational and memory costs. In order to improve scalability of GPs, approximate posterior inference is frequently employed, where a prominent class of approximation techniques is based on local GP experts. However, local-expert techniques proposed so far are either not well-principled, come with limited approximation guarantees, or lead to intractable models. In this paper, we introduce deep structured mixtures of GP experts, a stochastic process model which i) allows exact posterior inference, ii) has attractive computational and memory costs, and iii) when used as GP approximation, captures predictive uncertainties consistently better than previous expert-based approximations. In a variety of experiments, we show that deep structured mixtures have a low approximation error and often perform competitive or outperform prior work.
    Originele taal-2Engels
    TitelInternational Conference on Artificial Intelligence and Statistics (AISTATS)
    UitgeverijProceedings of Machine Learning Research
    StatusGeaccepteerd/In druk - 26 apr 2020
    Evenement23rd International Conference on Artificial Intelligence and Statistics - Palermo, Italië
    Duur: 3 jun 20205 jun 2020


    Congres23rd International Conference on Artificial Intelligence and Statistics
    Verkorte titelAISTATS 2020

    Bibliografische nota

    AISTATS 2020, in press


    • cs.LG
    • stat.ML

    Vingerafdruk Duik in de onderzoeksthema's van 'Deep Structured Mixtures of Gaussian Processes'. Samen vormen ze een unieke vingerafdruk.

    Citeer dit