Support vector learning for fuzzy rule-based classification systems.ppt
文本预览下载声明
Support vector learning for fuzzy rule-based classification systems Author: Zhong-dong Wu, Wei-xin Xie. Journal: Proceedings of the Fifth International Conference on Computational Intelligence and Multimedia Applications (ICCIMA’03) Report: C.C. Yen Clustering (1/3) 資料分群(data clustering)或是分群演算法(clustering algorithms)是一種將資料分類成群的方法,其主要的目的乃在於找出資料中較相似的幾個群聚(clusters),並找出各個群聚的代表點,稱為中心點(centroids)或是原型(prototypes)。使用這些中心點來代表原先大量的資料點,就可以達到兩個基本目標: 降低計算量 資料壓縮 Clustering (2/3) 一般而言,分群法可以大致歸為兩大類: 階層式分群法(hierarchical clustering):群數(number of clusters)可以由大變小,或是由小變大,來進群聚的合併或分裂,最後再選取最佳的群數。 分割式分群法(partitional clustering):先指定群數後,再用一套疊代的數學運算法,找出最佳的分群方式以及相關的群中心。 Clustering (3/3) 所有的分群法都有相似的流程,大略可歸納為下列三點: 收集資料 使用某種方法進行分群 測試分群結果 檢測分群結果,如果未達預期效果,則回到步驟二,再一次進行分群 Characteristic 它是如何分群的? 為何它要使用Positive–Definite Fuzzy Classifiers (PDFC) method? 是不是比較好? 還是在哪些情境下才能使用該篇論文的方法? 它套用哪些well-know method? 為何他要選用Mercer Kernel? 有哪些好處? Introduction 本篇講述利用fuzzy學習(判斷)input value應該要放再哪一群來做分類之系統. 作者發明在Binary Fuzzy Classifier底下使用positive-definite function成為一個Positive–Definite Fuzzy Classifiers (PDFC) method. go-back p5 Case-study Iris data set來說明本篇學習系統的效率及好處. Outline Introduction to fuzzy Introduction to FCMC Case study Simulate for IRIS Data Set Fuzzy http://buri.cool.ne.jp/2004/2483.jpg 馬 /SmartScience/Popa/Vol3-4.html 豬 /050406/47/1o2vw.html 金城武 goback-p16 go-back p6 Fuzzy cluster Fuzzy clustering is the part of the fuzzy data analysis that comprises two or more different area. Analysis of fuzzy data Analysis of usual (crisp) data with the help of fuzzy techniques Cluster analysis Aim of a cluster analysis is to partition a given set of data or objects into clusters (group, classes) Homogeneity within the clusters Heterogeneity between clusters Fuzzy-unsuitable Not always well suited for real time applications. Not always well suited for precision-instrument applications. Fuzzy c-means(1/2) FCM is a method of clustering which allows one piece of data to b
显示全部