A Tutorial on Bayesian Optimization of Expensive Cost Functions, with Application to Active User Modeling and Hierarchical Reinforcement Learning

ID
TR-2009-23
Authors
Eric Brochu, Mike Cora and Nando de Freitas
Publishing date
November 16, 2009
Length
50 pages
Abstract
We present a tutorial on Bayesian optimization, a method of finding the maximum of expensive cost functions. Bayesian optimization employs the Bayesian technique of setting a prior over the objective function and combining it with evidence to get a posterior function. This permits a utility-based selection of the next observation to make on the objective function, which must take into account both exploration (sampling from areas of high uncertainty) and exploitation (sampling areas likely to offer improvement over the current best observation). We also present two detailed extensions of Bayesian optimization, with experiments -- active user modelling with preferences, and hierarchical reinforcement learning. While the most common prior for Bayesian optimization is a Gaussian process, we also present random forests as an example of an alternative prior.