Learn from Global Correlations: Enhancing Evolutionary Algorithm via Spectral GNN
arXiv:2412.17629v4 Announce Type: replace-cross Abstract: Evolutionary algorithms (EAs) simulate natural selection but have two main limitations: (1) they rarely update individuals based on global correlations, limiting comprehensive learning; (2) they struggle with balancing exploration and exploitation, where excessive exploitation causes…
