GPSAF: A Generalized Probabilistic Surrogate-Assisted Framework for Constrained Single- and Multi-objective Optimization

• DOI: 10.48550/arxiv.2204.04054
• Publication Date: 2022-01-01
• Authors: Julian Blank; Kalyanmoy Deb
• Keywords: Surrogate model; Engineering optimization

Significant effort has been made to solve computationally expensive optimization problems in the past two decades, and various methods incorporating surrogates into have proposed. Most research focuses on either exploiting surrogate by defining a utility problem or customizing an existing method use one multiple approximation models. However, only little attention paid generic concepts applicable different types of algorithms simultaneously. Thus this paper proposes generalized probabilistic surrogate-assisted framework (GPSAF), broad category unconstrained constrained, single- multi-objective algorithms. The idea is based assisting method. assistance distinct phases, facilitating exploration another surrogates. exploitation are automatically balanced performing knockout tournament among clusters solutions. A study well-known population-based conducted with without proposed maximum solution evaluation budget 300 less. results indicate effectiveness applying GPSAF algorithm competitiveness other