HTSD-Framework for Evolutionary Algorithms

HTSD: A historical trajectory subspace decomposition guided framework for evolutionary algorithms

A drop-in search-enhancement module for population-based meta-heuristics. HTSD stores elite historical trajectories within a sliding window, applies SVD/PCA to extract dominant search directions, and decomposes the search space into a trend subspace and its orthogonal complement. It exploits along historically promising directions while exploring under-sampled orthogonal regions. A stagnation-aware controller adaptively regulates the exploitation and exploration weights.