Bayesian Optimization in a Billion Dimensions via Random Embeddings

Bayesian optimization techniques have been successfully applied to robotics, planning, sensor placement, recommendation, advertising, intelligent user interfaces and automatic algorithm configuration. Despite these successes, the approach is restricted to problems of moderate dimension, and several workshops on Bayesian optimization have identified its scaling to high-dimensions as one of the holy grails of the field. In this paper, we introduce a novel random embedding idea to attack this problem. The resulting Random EMbedding Bayesian Optimization (REMBO) algorithm is very simple, has important invariance properties, and applies to domains with both categorical and continuous variables. We present a thorough theoretical analysis of REMBO, including regret bounds that only depend on the problem’s intrinsic dimensionality. Empirical results confirm that REMBO can effectively solve problems with billions of dimensions, provided the intrinsic dimensionality is low. They also sho...

• Type: Journal
• Year: 2016
• Where: JAIR
• Authors: Ziyu Wang, Frank Hutter, Masrour Zoghi, David Matheson, Nando de Freitas
• Keywords: JAIR 2016, Software Engineering