Bayesian Optimization and Decision Making

Recent years have brought substantial advances in sequential decision making under uncertainty. These advances have occurred in many different communities, including several subfields of computer science, statistics, and electrical/mechanical/chemical engineering. While these communities are essentially trying to solve the same problem, they develop rather independently, using different terminology: Bayesian optimization, experimental design, bandits, active sensing, personalized recommender systems, automatic algorithm configuration, reinforcement learning, and so on. Some communities focus more on theoretical aspects while others' expertise is on real-world applications. This workshop aims to bring researchers from these communities together to facilitate cross-fertilization by discussing challenges, findings, and sharing data. This workshop follows last year's NIPS workshop "Bayesian optimization, experimental design and bandits: Theory and applications", one of the most-attended workshops in 2011. This year we plan to focus somewhat more on real-world applications, to bridge the gap between theory and practice. Specifically, we plan to have a panel discussion on real-world and industrial applications of Bayesian optimization and an increased focus on real-world applications in the invited talks (covering hyperparameter tuning, configuration of algorithms for solving hard combinatorial problems, energy optimization, and optimization of MCMC). Similar to last year, we expect to highlight the most beneficial research directions and unify the whole community by setting up this workshop.