English  |  正體中文  |  简体中文  |  Items with full text/Total items : 49378/84106 (59%)
Visitors : 7382509      Online Users : 54
RC Version 7.0 © Powered By DSPACE, MIT. Enhanced by NTU Library & TKU Library IR team.
Scope Tips:
  • please add "double quotation mark" for query phrases to get precise results
  • please goto advance search for comprehansive author search
  • Adv. Search
    HomeLoginUploadHelpAboutAdminister Goto mobile version
    Please use this identifier to cite or link to this item: http://tkuir.lib.tku.edu.tw:8080/dspace/handle/987654321/35777


    Title: Pso-based evolutionary learning : system design and applications
    Other Titles: 粒子群演算法為基礎的演化式學習 : 系統設計及其應用
    Authors: 陳慶逸;Chen, Ching-yi
    Contributors: 淡江大學電機工程學系博士班
    余繁;Ye, Fun
    Keywords: 粒子群演算法;群聚分析;群聚驗證;向量量化;類神經網路;Particle Swarm Optimization;Cluster analysis;Cluster Validity;Vector Quantization;Fuzzy c-means;Neural Networks
    Date: 2006
    Issue Date: 2010-01-11 07:08:31 (UTC+8)
    Abstract: 本論文回顧了演化式計算的重要方法和技術,從而探討粒子群演算法(PSO)及其在資料探勘、影像壓縮和類神經網路的應用。針對不同的處理問題,我們分析並結合不同的輔助機制來設計粒子群演算法的學習架構,以期得到有效的PSO系統應用平台(framework)。粒子群演算法是主要的演化式計算技術之一,它應用生物群體透過簡單模仿和跟隨其他個體而產生(emergent)系統自我組織和演化的概念,發展出最佳化演算法。PSO利用模仿群體性生物的社會行為(social-only model)取向和個體認知(cognition-only model)取向的機率性選擇來搜尋高維度問題空間的最佳解,其演算法相當簡單快速,解題原理又可以有效避開區域最小值,對於多模態最佳解(multi-modal)問題提供了一個相當良好的解決方案。在論文的第一部分,我們引進兩種以PSO為基礎的群聚分析演算法。首先我們提出一個結合指數型態距離與K-means運算法則的分割式群聚分析架構,它經由使用者在預設群聚數目的條件下產生最佳的分群結果。第二種PSO群聚分析演算法則是整合群聚驗證方法來得到資料探勘問題的最佳群聚數目以及群聚中心,以達到自動分群的目的。論文的第二部分在於提出一個應用在影像壓縮的模糊PSO向量量化器,該方法透過PSO參數學習和模糊推論生成影像向量的最佳化碼簿。相較於傳統的LBG方法,我們所提的架構更具有效性以及強健性。論文的最後一個部分致力於PSO在放射狀基底函數(RBF)神經網路的應用。針對網路的隱藏層節點與權重等參數,我們先使用正規化型式的Fuzzy c-means演算法(NFCM)進行粗略式(coarse-level)的結構鑑別,再經由結合遞迴式最小平方法則的PSO演算法作細部調整(fine tuning-level)的訓練;此一創新方法除了能以極小數量的PSO族群來達到實現RBF神經網路訓練的目的之外,其建模的性能和效率表現也得到很大的提升。本論文主要貢獻在於提出粒子群演算法的系統化的學習架構及其在工程設計最佳化問題領域的應用;基於此一泛用架構,未來我們得以快速發展出各種可靠的、高效能的工程最佳化系統。
    The new paradigm of Swarm Intelligence, called Particle Swarm Optimization (PSO), is one of the well-known evolutionary computation techniques, which can be considered as an efficient tool to find near optimal solution in a searching space. Especially, PSO is a useful method when the problems to be solved are high-dimensional, nonlinear or some specific information is unavailable. PSO combines the social-only model and the cognition-only model to select the adjustable parameters to approach optimal solution, its main advantage is its rapid convergence and small computational requirements, which make it a good candidate for solving optimization problems. In this dissertation, the efficient, robust, and flexible PSO algorithms are proposed to generate some artificial intelligence system in solving some applications, such as cluster analysis, image processing, and neural network training.
    The first task of this dissertation introduces two types of PSO clustering applications. The first one is given in advance the optimal number of clusters by manual manipulation, and then the PSO is applied to achieve the optimal clustering results. The other one is to use PSO algorithm that includes the cluster validity measure to automatically determine the true number of the cluster centers, and then to extract real cluster centers and to make a good classification.
    The second task of this dissertation is to develop an evolutional fuzzy particle swarm optimization (FPSO) learning algorithm to automatically extract the near-optimum codebook of vector quantization (VQ) for carrying on image compression. Based on the adaptive learning scheme of the PSO and the flexible membership function of the fuzzy inference system, the dissertation also demonstrates the advance of the FPSOVQ-based image compression system.
    The last issue of this dissertation is focuses on the topic of radial basis function networks (RBFNs) learning. An innovative hybrid recursive particle swarm optimization (HRPSO) learning algorithm with normalized fuzzy c-mean (NFCM) clustering is proposed to generate radial basis function networks (RBFNs) modeling system with small numbers of descriptive radial basis functions (RBFs) for fast approximating two complex and nonlinear functions.
    Appears in Collections:[電機工程學系暨研究所] 學位論文

    Files in This Item:

    File SizeFormat
    0KbUnknown298View/Open

    All items in 機構典藏 are protected by copyright, with all rights reserved.


    DSpace Software Copyright © 2002-2004  MIT &  Hewlett-Packard  /   Enhanced by   NTU Library & TKU Library IR teams. Copyright ©   - Feedback