合作式學習為基礎之混合型進化演算法在模糊類神經系統設計及多漏斗函數最佳化的應用

Abstract

在這篇論文中，我們主要在探討的是進化型演算法的合作學習機制。本論文中所探討的三種進化型演算法包括：基因演算法、粒子群聚最佳化演算法、以及自適應共變異數矩陣演化策略。在基因演算法的改良上，我們提出了族群式的共生演化概念，使得基因演算法可以將解空間分割成數個子空間，且在每個子空間中分別得去探索最佳解。我們也在合作式的粒子群聚演算法中提出了一個可分割度的偵測方法，以便將不可分割之變數置入同一族群中演化。至於關於自適應共變異數矩陣演化策略的改良，本論文提出了一個基於均值移動的平行運算機制，使得我們可以平行地在解空間中提供多個自適應共變異數矩陣演化策略學習器來探索解空間中的不同區域。論文的內容包括了將進化型演算法套用在模糊類神經系統上之架構和參數學習、演算法上的改良、平行運算機制以及結合兩種演算法優點的混合型演算法的研究。

In this dissertation, we mainly focus on researching the cooperative behavior of evolutionary algorithms. Algorithms discussed in this dissertation include genetic algorithm (GA), particle swarm optimization (PSO) and evolution strategy with covariance matrix adaptation (CMA-ES). The modification of genetic algorithm (GA) is done by introducing the group-based symbiotic evolution (GSE) technique which enables genetic algorithm (GA) to partition the search space into smaller subspaces and explore each smaller subspace by a separate agent to alleviate the curse of dimensionality. We also propose a separability detection method based on covariance matrix adaption mechanism into the cooperative particle swarm optimization (CPSO) to locate non-separable variables into the same swarm. As to the research of evolution strategy with covariance matrix adaptation (CMA-ES), we introduce the mean shift procedure which allows us to apply multiple CMA-ES instances to explore different parts of the search space in parallel. The scope of this dissertation includes how to implement evolutionary algorithms on neural-fuzzy systems, the improvement of algorithms, parallel computing and the emergence of two algorithms