Taboo Niches for Global Optimization via Simplex Search

摘要

This paper issues a novel Taboo Niches Global Optimization (TNGO) approach for solving continuous nonlinear global optimum problems by Simplex Search. Here niche is a subspace around a local optimum. The hybrid technique involves dual search stages, namely, random uniform global searching and local niche searching. And then, the searched niche is tabooed. Repeat above stages until all niches are tabooed one by one. Local searching is only once for each niche. Taboo niches ratio is used as an effective global convergence criterion. Classical Benchmark multimodal functions were used as for testing the approach and comparing with other global optimum algorithms mentioned in recent literature. The results showed its superiority.