A Generative Adversarial Network Guided Evolutionary Algorithm for Large-scale Sparse Multiobjective Optimization

In recent decades, great progress has been made in learnable multiobjective evolutionary algorithms (MOEAs) in the field of evolutionary computations. However, existing learnable MOEAs have not been equipped with powerful strategies for addressing the grand series associated with sparse large-scale multiobjective optimization problems (sparse LSMOPs), which include the curse of dimensionality and unknown sparsity characteristics. This work proposes a generative adversarial network (GAN)-guided evolutionary algorithm for solving sparse LSMOPs. GAN-aided offspring generation is adopted at each generation to generate high-quality sparse offspring solutions to improve the search performance, owing to the GAN′s powerful learning and generative capabilities. Specifically, random interpolation and discretization strategies are utilized to prevent mode collapse and falling into local optima, thereby generating promising sparse offspring solutions. The experimental results on both benchmark and real-world problems verify the superior performance of the proposed algorithm compared with the state-of-the-art evolutionary algorithms.