具吞噬作用之免疫演算法應用於多目標優化問題

Modified Immune Algorithm with Phagocytosis and Its Application in Multi-objective Optimization

于宗民(Tsung-Min Yu)

指導教授 ： 林志民

元智大學/電機通訊學院/電機工程學系/碩士(2013年)

https://doi.org/10.6838/YZU.2013.00033

摘要

本論文基於免疫演算法(immune algorithm)，提出了具吞噬作用之免疫演算法 (modified Immune Algorithm with Phagocytosis, PIA) 及具吞噬作用之多目標免疫演算法(MOPIA)及其應用。免疫演算法起源於生物免疫系統 (immune system)，本文提出之演算法加入以同步擾動隨機近似(Simultaneous perturbation stochastic approximation，SPSA)實現之吞噬作用，此外，基於抗獨特型抗體(anti-idiotype)的結構改變其突變機制，使得提出之PIA更符合自然的生物免疫系統，並基於李亞普諾夫定理(Lyapunov theorm)，藉由選擇適當的SPSA之學習步伐長度以確保PIA之收斂能力並加快優化速度。我們利用帕累托(Pareto)的概念將具吞噬作用之免疫演算法修正並應用於多目標優化問題(multi-objective optimization problem, MOP)。最後，我們使用PIA應用於陣列天線優化(pattern array)以及使用MOPIA去訓練RFNN並以非線性系統鑑別驗證MOPIA之性能。

關鍵字

免疫演算法 ； 吞噬作用 ； 抗獨特型抗體 ； 李亞普諾夫定理 ； 帕累托 ； 陣列天線優化 ； 非線性系統辨識

並列摘要

In this thesis, we propose the modified Immune Algorithm with phagocytosis for multi-objective optimization problem (MOPIA). At first, the modified Immune Algorithm with phagocytosis (PIA) is developed for single objective optimization problem. The proposed PIA algorithm participate phagocytosis, besides, alter the mutation strategy based on the structure of anti-idiotype, hence, PIA is resemble to nature biological immune system. In addition, the learning step length is derived by Lyapunov stability approach to guarantee the convergence and faster evolution. Based on the concept of Pareto, the PIA is modified for solving multi-objective optimization problem, called MOPIA. Finally, we use PIA application in antenna array optimization and MOPIA for training and optimizing the rule numbers of RFNN in nonlinear dynamic system to demonstrate the effectiveness.

並列關鍵字

Immune algorithm ； Phagocytosis ； Anti-idiotype ； Lyapunov theorem ； Pareto ； Antenna array optimization ； Nonlinear system identification

參考文獻

[2] A. Abraham and L. Jain, “Evolutionary Multi-objective Optimization: Theoretical Advances and Applications,” Springer, 2005.

[3] F. M. Burnet, The Colonal Selection Theory of Acquired Immunity, Cambridge University Presss, 1959.

[4] G. B. Bezerra, L. N. de Castro, and F. J. V. Zuben, “A hierachical immune network applied to gene expression data,” in Proc. 3rd Int. Conf. Artif. Immune Syst., pp. 14–27, 2004.

[5] C. A. Coello and N. C. Cortes, “Solving Multi-objective Optimization Problems Using an Artificial Immune System,” Genetic Programming and Evolvable Machines, Vol. 6, pp. 163–190, 2005.

[6] De Castro L. N and Von Zuben F. J, “The clonal selection algorithm with engineering applications,” Workshop Proceedings of GECCO, pp. 31-37, 2000.