webinar register page

Webinar banner
On MCMC for variationally sparse Gaussian process: A pseudo-marginal approach - Sara Wade and Karla Monterrubio-Gómez
Gaussian processes (GPs) are frequently used in machine learning and statistics to construct powerful models. However, when employing GPs in practice, important considerations must be made, regarding the high computational burden, approximation of the posterior, form of the covariance function and inference of its hyperparmeters. To address these issues, Hensman et al. (2015) combine variationally sparse GPs with Markov chain Monte Carlo (MCMC) to derive a scalable, flexible, and general framework for GP models. Nevertheless, the resulting approach requires intractable likelihood evaluations for many observation models. To bypass this problem, we propose a pseudo-marginal (PM) scheme that offers asymptotically exact inference as well computational gains through doubly stochastic estimators for the intractable likelihood and large datasets. In complex models, the advantages of the PM scheme are particularly evident, and we demonstrate this on a two-level GP regression model with a nonparametric covariance function to capture non-stationarity.

Dec 16, 2020 05:00 PM in Paris

Webinar logo
* Required information
Loading

By registering, I agree to the Privacy Statement and Terms of Service.

Register

Speakers

Sara Wade
Assistant Professor @University of Edinburgh
Dr. Sara Wade a Lecturer (Assistant Professor) in Statistics and Data Science at the University of Edinburgh and an elected officer of the International Society of Bayesian Analysis (ISBA). Her research is at the intersection of Bayesian nonparametrics and machine learning, with a focus on flexible nonparametric priors and efficient inference for complex data, with applications in biomedical studies, including the diagnosis and progression of dementia based on clinical, imaging and biological data. In 2018, she chaired the 4th edition of BAYSM, a conference dedicated to early-career researchers in Bayesian analysis. She was awarded Best Doctoral Thesis in Statistics 2014 by the Italian Statistical Society. https://www.maths.ed.ac.uk/~swade/
Karla Monterrubio-Gómez
Postdoctoral Research Associate in Biomedical Data Science @MRC Human Genetics Unit at the University of Edinburgh
Karla Monterrubio-Gómez is a Postdoctoral Research Associate in Biomedical Data Science at MRC Human Genetics Unit at the University of Edinburgh. After receiving her Ph.D. in Statistics from the University of Warwick, she held a Research Associate position at PROWLER.io focusing on computationally efficient non-stationary Gaussian process models. Her research interests lie at the interface of Bayesian statistics and computer science developing scalable algorithms for biomedical and environmental applications.