Henry Moss - MUMBO: MUlti-task Max-value Bayesian Optimization
|Starts:||12:00 5 Nov 2019|
|Ends:||13:00 5 Nov 2019|
|What is it:||Seminar|
|Organiser:||Department of Mathematics|
|Who is it for:||University staff, External researchers, Current University students|
Join us for this research seminar, part of the SQUIDS (Statistics, quantification of uncertainty, inverse problems and data science) seminar series.
Abstract: We propose MUMBO, the first high-performing yet computationally efficient acquisition function for multi-task Bayesian optimization. Here, the challenge is to perform efficient optimization by evaluating low-cost functions somehow related to our true target function, a broad class of problems including the popular task of multi-fidelity optimization. However, while information-theoretic acquisition functions are known to provide state-of-the-art Bayesian optimization, existing implementations for multi-task scenarios have prohibitive computational requirements. Previous acquisition functions have therefore been suitable only for problems with both low-dimensional parameter spaces and function query costs sufficiently large to overshadow very significant optimization overheads. In this work, we derive a novel multi-task version of the max-value entropy search of Wang et al 2017, delivering low-cost and robust performance across classic optimization challenges and multi-task hyper-parameter tuning tasks. Our approach is scalable and efficient, allowing multi-task Bayesian optimization to be deployed in problems with rich parameter and fidelity spaces.
Organisation: Lancaster University
Travel and Contact Information
Frank Adams 1
Alan Turing Building