Evolutionary Generative Optimization: Towards Fully Data-Driven Evolutionary Optimization via Generative Learning

Abstract: Recent advances in data-driven evolutionary algorithms (EAs) have demonstrated the potential of leveraging data to improve optimization accuracy and adaptability. Nevertheless, most existing approaches remain dependent on handcrafted heuristics, which limits their generality and automation. To address this challenge, we propose Evolutionary Generative Optimization (EvoGO), a fully data-driven framework empowered by generative learning. EvoGO streamlines the evolutionary optimization process into three stages: data preparation, model training, and population generation. The data preparation stage constructs a pairwise dataset to enrich training diversity without incurring additional evaluation costs. During model training, a tailored generative model learns to transform inferior solutions into superior ones. In the population generation stage, EvoGO replaces traditional reproduction operators with a scalable and parallelizable generative mechanism. Extensive experiments on numerical benchmarks, classical control problems, and high-dimensional robotic tasks demonstrate that EvoGO consistently converges within merely 10 generations and significantly outperforms a wide spectrum of optimization approaches, including traditional EAs, Bayesian optimization, and reinforcement learning based methods. Source code will be made publicly available.