BOSH : Bayesian Optimization by Sampling Hierarchically

Moss, Henry B. and Leslie, David S. and Rayson, Paul (2020) BOSH : Bayesian Optimization by Sampling Hierarchically. In: Workshop on Real World Experiment Design and Active Learning at ICML 2020, 2020-07-13 - 2020-07-18.

Text (2007.00939v1)
2007.00939v1.pdf - Published Version
Download (984kB)

Abstract

Deployments of Bayesian Optimization (BO) for functions with stochastic evaluations, such as parameter tuning via cross validation and simulation optimization, typically optimize an average of a fixed set of noisy realizations of the objective function. However, disregarding the true objective function in this manner finds a high-precision optimum of the wrong function. To solve this problem, we propose Bayesian Optimization by Sampling Hierarchically (BOSH), a novel BO routine pairing a hierarchical Gaussian process with an information-theoretic framework to generate a growing pool of realizations as the optimization progresses. We demonstrate that BOSH provides more efficient and higher-precision optimization than standard BO across synthetic benchmarks, simulation optimization, reinforcement learning and hyper-parameter tuning tasks.