University of Cambridge > > Machine learning in Physics, Chemistry and Materials discussion group (MLDG) > Introduction to Gaussian processes and Modulated Bayesian Optmisation

Introduction to Gaussian processes and Modulated Bayesian Optmisation

Add to your list(s) Download to your calendar using vCal

If you have a question about this talk, please contact Bingqing Cheng .

  • Abstract

This talk will be divided into two parts. In the first part I will provide an introduction to stochastic processes and in specific Gaussian processes. We will see how these mathematically simple objects can be used to specify distributions over the space of functions. This allows for Bayesian inference over function providing a principled treatment of uncertainty.

The second part of the talk will focus on sequential decision making under uncertainty. In specific we will discuss Bayesian Optmisation (BO) where the goal is to find the extremum of an explicitly unknown function. By placing a GP-prior as a surrogate to the unknown function we can formulate optimisation as a search strategy. However, the BO methodology suffers from the fact that the modelling task is distinct from the search. This often leads to sub-optimal decisions when structures irrelevant for the optima have been observed. In this talk I will describe recent worked called “Modulating Surrogates for Bayesian Optmisation”. The goal of this paper is to provide a surrogate model that allows only structures that are relevant for the search to be modelled.

  • bio

Dr. Carl Henrik Ek is a recently appointed senior lecturer in Computer Science at the University of Cambridge. Prior to Cambridge he worked at the University of Bristol. His research focuses on Bayesian non-parametrics and stochastic processes. In specific he is interested in building models for scenarios where data is scares and uncertain for critical applications that requires statistical models that can be interrogated and interpret-able.

This talk is part of the Machine learning in Physics, Chemistry and Materials discussion group (MLDG) series.

Tell a friend about this talk:

This talk is included in these lists:

Note that ex-directory lists are not shown.


© 2006-2020, University of Cambridge. Contact Us | Help and Documentation | Privacy and Publicity