使用遺傳基因法和倒傳遞法來訓練模糊神經網路做函數的近似

Abstract

在本論文中,我們使用遺傳基因法和倒傳遞法來訓練模糊神經網路做函數 的近似.遺傳基因法和倒傳遞法是用來調模糊神經網路的參數.傳統條模糊 神經網路參數的方法(倒傳遞法)有個弱點就是必須依靠初使條件(連線出 始化或非連線出始化).為了解決這問題我們使用遺傳基因法和到傳遞法來 調模糊神經網路的參數. In this thesis, we present a fuzzy neural network system for a function approximation, which is trained by genetic algorithms and back propagation.The genetic algorithms and back propagation are used for tuning the FNN model parameters. The conventional method(back propagation) has a weak point that the structure of FNN model depends on initial conditions(on-line or of-line initialization).In order to solve this problem this paper proposes a tuning method for the FNN model by GA and BP.