Evolutionary Algorithms for Sparse Large‐Scale Multi‐Objective Optimization

In addition to developing multi‐objective evolutionary algorithms (MOEAs) for solving general large scale multi‐objective optimization problems (LSMOPs), it is reasonable to improve the convergence speed by involving problem‐specific information in offspring generation, such as the gradients used in mathematical programming methods. This chapter presents the first three MOEAs for sparse large‐scale multi‐objectiveoptimization. The most significant feature of these MOEAs is that they search for the sparse Pareto optimal solutions of specific LSMOPs by customizing effective strategies, such as the population initialization strategy in SparseEA, the Pareto optimal subspace learning strategy in MOEA/PSL, and the evolutionary frequent pattern mining method in PM‐MOEA. Due to the wide application scenarios of sparse large‐scale multi‐objective optimization, it appears as an emerging topic in evolutionary multi‐objective optimization, and an increasing number of new methods are being proposed.