English

Mercer Features for Efficient Combinatorial Bayesian Optimization

Machine Learning 2022-02-07 v1 Artificial Intelligence

Abstract

Bayesian optimization (BO) is an efficient framework for solving black-box optimization problems with expensive function evaluations. This paper addresses the BO problem setting for combinatorial spaces (e.g., sequences and graphs) that occurs naturally in science and engineering applications. A prototypical example is molecular optimization guided by expensive experiments. The key challenge is to balance the complexity of statistical models and tractability of search to select combinatorial structures for evaluation. In this paper, we propose an efficient approach referred as Mercer Features for Combinatorial Bayesian Optimization (MerCBO). The key idea behind MerCBO is to provide explicit feature maps for diffusion kernels over discrete objects by exploiting the structure of their combinatorial graph representation. These Mercer features combined with Thompson sampling as the acquisition function allows us to employ tractable solvers to find next structures for evaluation. Experiments on diverse real-world benchmarks demonstrate that MerCBO performs similarly or better than prior methods. The source code is available at https://github.com/aryandeshwal/MerCBO .

Keywords

Cite

@article{arxiv.2012.07762,
  title  = {Mercer Features for Efficient Combinatorial Bayesian Optimization},
  author = {Aryan Deshwal and Syrine Belakaria and Janardhan Rao Doppa},
  journal= {arXiv preprint arXiv:2012.07762},
  year   = {2022}
}

Comments

10 pages, 5 figures, Accepted at AAAI conference 2021