以函數鏈結為基礎之類神經模糊網路及其應用

Abstract

本篇論文提出一以函數鏈結為基礎之類神經模糊網路及其相關學習演算法。此類神經模糊網路採用函數鏈結類神經網路當作模糊法則的後件部。此後件部是輸入變數的非線性組合，它是利用函數展開的方式，能在高維度的輸入空間中提供良好的非線性決策能力，因此，可使網路輸出更具體且更逼近目標輸出。本論文主要為三大部分。第一部份將詳細介紹以函數鏈結為基礎之類神經模糊網路及其線上學習演算法。此演算法包含架構學習及參數學習，架構學習是藉由熵的量測決定是否要增長一個新的法則，參數學習是使用倒傳遞演算法調整網路上的所有參數。由於倒傳遞演算法常常會得到局部最佳解。因此，在第二部份中，我們提出一改良式差分進化演算法，所提出的演算法與傳統差分進化演算法是不同的，在於我們使用一有效的搜尋機制使得每條個體能更新在目前最佳解和亂數搜尋解之間，並採用以群為基底的突變方式以提高個體間彼此的差異性。但以上的進化演算法，無法決定該使用多少法則數。因此，在第三部份中，我們提出一以法則為基礎的共生差分進化演算法。此演算法是利用多個子族群進行進化，每個子族群的個體代表每條模糊法則，且每個子族群能各自進化。此外，這演算法也能自動決定子族群數，並最佳化網路上的所有參數。最後，我們將與其他方法比較，以證實所提出的網路架構及其相關演算法之有效性。

This dissertation proposes a functional-link-based neuro-fuzzy network (FLNFN) and its related learning algorithms. The proposed FLNFN model uses a functional link neural network to the consequent part of the fuzzy rules. The consequent part uses a nonlinear functional expansion to form arbitrarily complex decision boundaries. Thus, the local properties of the consequent part in the FLNFN model enable a nonlinear combination of input variables to be approximated more effectively. This dissertation consists of three major parts. In the first part, the FLNFN model and an online learning are presented. The online learning algorithm consists of structure learning and parameter learning. The structure learning depends on the entropy measure to determine the number of fuzzy rules. The parameter learning, based on back-propagation, can adjust the shape of the membership function and the corresponding weights of the consequent part. Unfortunately, the back-propagation learning algorithm may reach the local minima very quickly. Therefore, a modified differential evolution (MODE) is presented to optimize the FLNFN parameters in the second part. The proposed MODE learning algorithm differs from the traditional differential evolution. The MODE adopts a method to effectively search between the best individual and randomly chosen individuals, and the MODE also provides a cluster-based mutation scheme, which maintains useful diversity in the population to increase the search capability. But, the aforementioned algorithm cannot determine how many rules to be used. Therefore, a rule-based symbiotic modified differential evolution (RSMODE) is proposed for the FLNFN model in the third part. The RSMODE adopts a multi-subpopulation scheme that uses each individual represents a single fuzzy rule and each individual in each subpopulation evolves separately. Furthermore, the proposed RSMODE learning algorithm can also determine the number of rule-based subpopulation and adjust the FLNFN parameters. Finally, the proposed FLNFN model and its related learning algorithms are applied in various control problems. Results of this dissertation demonstrate the effectiveness of the proposed methods.