A twofold infill criterion-driven heterogeneous ensemble surrogate-assisted evolutionary algorithm for computationally expensive problems

Abstract Over the past decade, surrogate-assisted evolutionary algorithms (SAEAs) have received considerable attention in solving expensive optimization problems from both academia and industry. In SAEAs, model management plays a key role in the all-encompassing use of surrogates and data. Constructing sophisticated management strategies and high-fidelity surrogates, nevertheless, remains an urgent and foundational task. In this context, we propose a twofold infill criterion-driven heterogeneous ensemble surrogate-assisted neighborhood field optimization algorithm (HESNFO). The proposed algorithm takes into account both the diversity and accuracy of surrogates to speed up the optimization process. A twofold infill criterion is presented to strike a balance between exploration and exploitation on the basis of updating the surrogates online. Meanwhile, to enhance the diversity of surrogates, a heterogeneous ensemble surrogate consisting of multiple radial basis function models with different architectures and different inputs has been built. Finally, our experimental results on several benchmark problems as well as an electromagnetic acoustic transducers optimization instance demonstrate that the proposed algorithm is superior.