Quantitative analysis based tuning law for convergence rate of Spiral Optimization
Kenichi Tamura, Keiichiro Yasuda · 2012
Recently, the authors proposed a new metaheuristics method for continuous optimization problems based on analogy of spiral phenomena in nature which is called Spiral Optimization. The focused spiral phenomena are spirals which are approximated to logarithmic spirals. The Spiral Optimization utilizes a feature of the logarithmic spirals for global optimization. The Spiral Optimization has two tuning parameters: the convergence rate r and the rotation angle θ. However, any tuning methods and policies for them were not studied and referred in our previous works. This paper especially focuses on the convergence rate r and quantitatively analyzes its properties. Furthermore, by using the analysis result, a simple the convergence rate tuning law which can match various computational conditions and problems is proposed.