Surrogate-assisted multi-objective evolutionary optimization with a multi-offspring method and two infill criteria

In this paper, we propose incorporating the surrogate-assisted multi-offspring method and surrogate-based infill points into a multi-objective evolutionary algorithm to solve high-dimensional computationally expensive problems. To enhance search efficiency and speed, multiple offspring are produced by the parent solutions. A hierarchical pre-screening criterion is proposed to select the surviving offspring and exactly evaluated offspring. The pre-screening criterion can maintain offspring diversity and superiority by using the non-dominated rank and reference vectors. Only a few offspring with good diversity and convergence are exactly evaluated in order to reduce the number of consumed function evaluations. Additionally, two types of surrogate-based infill points are used to further improve search efficiency. Pareto front model-based infill points are mainly used to enhance the exploration of sparse areas in the approximate Pareto front, while infill points from the surrogate-assisted local search are mainly used to accelerate the exploitation towards the real Pareto front. ZDT and DTLZ cases, with dimensions varying from 8 to 200, were adopted to test the performance of the proposed algorithm. Experi-mental results demonstrate the superiority of the proposed algorithm over the compared algorithms.

Automated summary

This paper proposes a multi-objective evolutionary algorithm that incorporates the surrogate-assisted multi-offspring method and surrogate-based infill points to solve high-dimensional computationally expensive problems. The algorithm produces multiple offspring to enhance search efficiency and speed, and uses a hierarchical pre-screening criterion to select surviving offspring and exactly evaluated offspring. Surrogate-based infill points are used to further improve search efficiency. Experimental results demonstrate the superiority of the proposed algorithm over compared algorithms.