Surrogate-based Global Optimization Methods for Expensive Black-Box Problems: Recent Advances and Future Challenges

Abstract

The great computational burden caused by complicated and unknown analysis restricts the use of simulation based optimization. In order to mitigate this challenge, surrogate-based global optimization methods have gained popularity for their capability in handling expensive black-box problems. This paper surveys the fundamental issues that arise in surrogate-based global optimization (SBGO) from a practitioner’s perspective, including highlighting concepts, methods, techniques as well as engineering applications. To provide a brief discussion on the issues involved, recent advances in design of experiments, surrogate modeling techniques, infill criteria and design space reduction are investigated. Future challenges and research is also analyzed and discussed.