A GAN-Assisted Bi-Stage Bayesian Co-Evolution Optimization Algorithm for High-dimensional Expensive Many-Objective Problems

Bayesian optimization (BO) struggles with data scarcity and poor scalability in high-dimensional expensive many-objective optimization problems (HEMaOPs). To address this, we propose a novel GAN-Assisted Bi-Stage Bayesian Optimization approach (GBB-CEO), a novel framework that synergizes generative adversarial networks (GANs) with Bayesian co-evolutionary search for data-driven optimization. The GAN module generates synthetic samples conditioned on promising regions identified by BO, while a co-evolutionary mechanism maintains two interacting populations: one explores the GAN’s latent space for diversity, and the other exploits BO’s probabilistic model for convergence. A bi-stage infilling strategy further enhances efficiency: early iterations prioritize exploration via Lp -norm-based candidate selection, later switching to a max-min distance criterion for Pareto refinement. Experiments on expensive multi/many-objective benchmarks show GBB-CEO outperforms three state-of-the-art surrogate-assisted algorithms, achieving superior convergence and diversity under limited evaluation budgets.