應用自我組織映射圖於群聚及視覺化之分析

A Study on Self-Organizing Map for Clustering Analysis and Visualization

謝淑玲(Shu-Ling Shieh) ； 廖宜恩(I-En Liao) ； 陳恆裕(Hun-Yu Chen)

《嶺東學報》 24期 (2008/12) Pp. 177-197

https://doi.org/10.29850/LTJ.200812.0009

摘要

自我組織映射圖網路(Self-Organizing Map, SOM)是一種非監督式學習網路(Unsupervised learning network)，主要目的是以映射方法將輸入資料轉換至映射圖上，讓自我組織映射圖中的神經元也能保有輸入樣本的拓撲結構。本研究對SOM演算法的神經元初始方式提出一個新方法，用以改善SOM的運算量，可使其執行時間大幅度的降低。最後，利用視覺化SOM的圖形，將SOM分群後的結果呈現在三維空間上，透過圖形的輔助，可使分群之後的結果更具有可信度。

關鍵字

分群法 ； 類神經網路 ； 自我組織映射圖 ； 資訊視覺化

並列摘要

The self-organizing map (SOM) is an unsupervised neural network which projects high-dimensional data onto a low-dimensional grid and preserving its 'topological' structure. In this paper, we propose an efficient self-organizing map method to improve the performance of the SOM. The proposed algorithm reduces the time complexity dramatically in finding the initial neurons. Finally, we use visualization methods to show the mapping from clustered results of SOM to a 3-D Euclidian space. By means of the assistance of graphic presentation, the outcome of SOM clustering process will achieve more reliability.

並列關鍵字

Clustering Methods ； Neural Network ； Self-Organizing Map ； Information Visualization

參考文獻

A. K. Jain,M. N. Murt,P. J. Flynn(1999).Data Clustering: A Review.ACM Computing Surveys.41(3)

A. Ultsch,H. P. Siemon(1990).Kohonen's self organizing feature maps for exploratory data analysis.Proc. INNC'90, Int. Neural Network Conf..(Proc. INNC'90, Int. Neural Network Conf.).

Alfred Ultsch(2004).U*-Matrix: a Tool to visualize Clusters in high dimensional Data.Technical report 36.(Technical report 36).,::CS Depratment, philipps-University Marburg.

UCI repository of machine learning databases

D. DeSieno(1988).Adding a conscience to competitive learning.IEEE ICNN.1,117-124.