群聚法(clustering)是一種非監督式學習(unsupervised learning)的分類方法，透過分析資料彼此的關係來達到分群的效果。支持向量群聚(support vector clustering)則是一種最新的群聚方法，它是從支持向量分類器(support vector machine)的理論所改良衍生出來，其基本的原理就是透過一個核心函數(kernel function)將資料投射到一個高維的特徵空間上(feature space)。而我們希望能在這空間中找到一個最小的球體來包含所有的資料，當這個球體投射回原來的空間時，就會形成許多組的輪廓(contours)，而這些輪廓就能夠將資料給區隔開來。誠如其它的群聚法，我們事先並不曉得最終會分成幾群出來，而且核心函數的參數值將會影響到輪廓的形狀也就是改變分群的結果。因此如何決定一個適當的參數值對於得到一個好的分群結果而言是相當困難且重要的。

　　為了減少這些不確定的因素而保留原有支持向量群聚的特性，在本論文中，我們提出了一套智慧型支持向量群聚(intelligent support vector clustering)的演算法，有別以往傳統的支持向量群聚，我們改採用了監督式學習(supervised learning)的方式。換言之，我們事先知道有K個類別，針對不同的類別我們各自去計算出K個不同的球體，而每一個球體都有其自己的核心函數參數。然而，由於不同類別的球體間可能會有重疊的現象，這將會導致錯誤分類的發生。因此，我們必須要決定一組最恰當的核心函數參數來儘量消弭這類重疊情況發生。本論文中，我們透過結合進化規劃法(evolutionary programming)及改良式渾沌最佳演算法(chaos optimal algorithm)的一種新全區域搜尋法――渾沌進化演算論(chaos evolutionary programming)來更有效率地得到此組最佳參數。

　　在實驗中，我們針對了兩個較大的分類問題來做探討。一是臉部表情辨識問題，二是乳房X光片的腫瘤偵測。在臉部表情辨識中，我們採用了著名的 Cohn-Kanade Face Expression資料庫，分別就多人來辨識高興、驚訝、生氣及難過四種表情做辨識。根據300多張影像的測試結果，我們所提出的新方法在對於生氣及難過這兩種表情上能有相當顯著的提升，而整體的辨識率亦有不錯的成果；另一方面，在腫瘤偵測上，我們針對緻密腺體一類，比較了四種不同的分類方法: 線性辨識分析(linear discriminant analysis)、倒傳遞類神經網路(back propagation neural network)、機率類神經網路(probabilistic neural network)和輻射基底類神經網路(radial basis function neural network)。實驗結果顯示我們的方法確時比其它方法有較高的辨識成效。

　　Clustering is an unsupervised learning classification method which has been utilized widely. Recently, a novel clustering method: the support vector clustering which is derived from the support vector machine was first proposed in 2000. Its basic concept is that data points are mapped into a high dimensional feature space by a kernel function. In the feature space, we look for the smallest sphere which encloses the image of the data. This sphere corresponds to a set of contours which enclose all the data points in the original input space. Like other clustering methods, we don’t know the final numbers of class in advance. Besides, the shape of contour will be influenced by the value of kernel function’s parameter, i.e. this also affects the clustering result. Therefore, it is an important and hard issue to decide a proper value of parameter for a good clustering result.

　　To keep properties of the support vector clustering but reduce the uncertainty of kernel function’s parameter, an intelligent support vector clustering algorithm is proposed. Unlike the traditional support vector clustering, we adopt a supervised learning approach. In other words, there will be K different spheres for K classes problem. However, because a confused classification may occur by the overlapping condition among distinct sphere, we should pick up a suitable combination of parameters to reduce overlapping conditions. Here, a new global search method: the chaos evolutionary programming which combines the evolutionary programming with the modified chaos optimal algorithm is used to gain optimal parameters with a higher efficiency.

　　Two classification problems: facial expression recognition and mass detection in mammograms are experimented with our new method. In facial expression recognition problem, we adopt famous Cohn-Kanade Face Expression data for classifying four expressions: happiness, surprise, anger and sadness. According to the result obtained from testing more than 300 images, our proposed method performs well especially in separating anger and sadness expressions. In mass detection, we focus on dense type and compare with four different classification methods: linear discriminant analysis(LDA), back propagation neural network(BPN), probabilistic neural network(PNN) and radial basis function neural network(RBF). Experimental results reveal that recognition rate through our method is higher than other methods.

[1] A. K. Jain, M. N. Murty, and P.J. Flynn, “Data Clustering: A Review,” ACM
Computing Surveys, vol. 31, no. 3, September 1999.
[2] V. L. Brailovsky, “A Porbabilistic Approach to Clustering,” Pattern
Recognition Letter, vol. 4, no. 4, pp. 193-198, April, 1991.
[3] C. T. Zahn, “Graph-theoretical Methods for Detecting and Describing Gestalt
Clusters,” IEEE Trans. on Computers, vol. C-20, pp. 68-86, 1971.
[4] A. Ben-Hur, H. T. Siegelmann, and V. N. Vapnik, “A Support Vector Clustering
Method,” in Proc. Int. Conf. Pattern Recognition, vol. 2, pp. 728–732,
2000.
[5] A. Ben-Hur, D. Hon, H. T. Siegelmann, and V. N. Vapnik, “Support Vector
Clustering,” Journal of Machine Learning Research, vol. 2, pp. 125-137,
2001.
[6] B. Boser, I. Guyon, and V. Vapnik, “A Training Algorithm for Optimal Margin
Classifiers,” Fifth Annual Workshop on Computational Learning Theory. ACM
Press, Pittsburgh, 1992.
[7] Bing-Yu Sun and De-Shuang Huang, “Support Vector Clustering for Multiclass
Classification Problems,” Congress on Evolutionary Computation, vol. 2, pp.
1480-1485, 2003.
[8] C. W. Hsu, “A Comparison of Methods for Multiclass Support Vector
Machines,”IEEE Trans. on Neural Network, vol. 13, pp. 415-425, 2002.
[9] J. Platt, “Fast Training of Support Vector Machines Using Sequential Minimal
Optimization,” Advance in Kernel Methods: Support Vector Learning, pp.
185:208, 1999.
[10] Jeen-Shing Wang and Jen-Chieh Chiang, “A Validity-Guided Support Vector
Clustering Algorithm for Identification of Optimal Cluster Configuration,”
IEEE International Conf. on System, Man and Cybernetics, 2004.
[11] Jung-Hsien Chiang and Pei-Yi Hao, “A New Kernel-Based Fuzzy Clustering
Approach: Support Vector Clustering with Cell Growing,” IEEE Trans. on
Fuzzy Systems, vol. 11, no. 4, August 2003.
[12] D.M.J. Tax and R.P.W. Duin., “Support Vector Domain Description,” Pattern
Recognition Letters, vol. 20, pp.1991–1999, 1999.
[13] J. H. Holland, Adaptation in Natural and Artificial Systems, Addison-Wesley,
Reading, MA, 1975.
[14] D. B. Fogel, Evolutionary Computation: Toward a New Philosophy of Machine
Intelligence, 2nd ed., IEEE press, Piscataway NJ, 1999
[15] Shu-Mei Guo, Leang S. Shieh, Guanrong Chen, and Norman P. Coleman,
“Observer-Type Kalman Innovation Filter for Uncertain Linear Systems,” IEEE
Trans. on Aerospace and Electronic Systems, vol. 37, no. 4, pp. 1406-1418,
October 2001.
[16] J. H. Halton, “On the Efficiency of Certain Quasi-random Sequences of
Points in Evaluating Multidimensional Integrals,” Numerische Mathematik,
vol. 2, p.84-90, 1960.
[17] B. Li and W. S. Jiang, “Optimizing Complex Functions by Chaos Search,”
Cybernetics and Systems, vol. 29, no. 4, pp. 409-419, June, 1998.
[18] Xuefeng F. Yan, Dezhao Z. Chen, and Shangxu X. Hu, “Chaos-genetic
Algorithms for Optimizing the Operating Conditions based on RBF-PLS Model,”
Computers and Chemical Engineering, vol. 27, pp. 1393-1404, October, 2003.
[19] Jianhua Yang, Vladimir Estivill-Castro, and Stephan K. Chalup, “Support
Vector Clustering through Proximity Graph Modeling,” in Proc. Int. Conf.
Neural Information Processing, vol. 2, 2002.
[20] Jaewook Lee and Daewon Lee, “An Improved Cluster Labeling Method for
Support Vector Clustering,” IEEE Trans. on Pattern Analysis and Machine
intelligence, vol. 27, no. 3, March 2005.
[21] A. Mehrabian, “Communication without Words,” Psychol. Today, vol2, pp.
53-56,1978.
[22] Yongsheng Gao, Maylor K. H. Leung, and Siu Cheung Hui, “Facial Expression
Recognition from Line-Based Caricatures,” IEEE Trans. on Systems, Man, and
Cybernetics, vol. 33, no. 3, pp. 407-412, May 2003.
[23] B. Fasel and Juergen Luettin, “Automatic Facial Expression Analysis: a
Survey,” Pattern Recognition, vol. 36, pp. 259-275, 2003.
[24] P. Ekman and W.V. Friesen, “Constants across Cultures in the Face and
Emotion,”J. Personality Social Psychol. vol. 17, no. 2, pp. 124-129, 1971.
[25] P. Ekman and W. V. Friesen, “The Facial Action Coding System: A Technique
for the Measurement of Facial Movement,” San Francisco: Consulting
Psychologists Press, 1978.
[26] G. Donato, M. S. Bartlett, J. C. Hager, P. Ekman, and T. J. Sejnowski,
“Classifying Facial Actions,” IEEE Trans. on Pattern Analysis and Machine
Intelligence, vol. 21, pp. 974-989, October 1999.
[27] Xue-Wen Chen and Thomas Huang, “Facial Expression Recognition: A
Clustering-based Approach,” Pattern Recognition Letters, vol. 24, pp.
1295-1302, 2003.
[28] Rein-Lien Hsu, Mohamed Abdel-Mottaleb, and Anil k. Jain, “Face Detection in
Color Image,” IEEE Trans. on Pattern Analysis and Machine intelligence,
vol. 24, no. 5, pp. 696-706, May 2002.
[29] Maja Pantic and Leon J.M. Rothkrantz, “Automatic Analysis of Facial
Expressions: The State of the Art,” IEEE Trans. on Pattern Analysis and
Machine Intelligence, vol. 22, no. 12, December 2000.
[30] Jyh-Yeong Chang and Jia-Lin Chen, “A Facial Expression Recognition System
Using Neural Networks,” International Joint Conference on Neural Networks,
vol. 5, pp. 3511-3516, July 1999.
[31] Ma, L. and Khorasani, K, “Facial Expression Recognition Using Constructive
Feedforward Neural Networks,” IEEE Transactions on Systems, Man and
Cybernetics, vol. 34, pp. 1588 – 1595, June 2004.
[32] Goudeaux, K, Tsuhan Chen, Shyue-Wu Wang, and Jen-Duo Liu, “Principal
Component Analysis for Facial Animation,” IEEE International Conference on
Acoustics, Speech, and Signal Processing, vol.3, pp. 1501 – 1504, May 2001.
[33] 吳明衛, “自動化臉部表情分析系統,” 成功大學資訊工程所, 2003.
[34] 潘奕安, “低解析度影像序列之自動化表情辨識系統,” 成功大學資訊工程所,
2004.
[35] 于南書, “最佳特徵選擇: 乳房X光片腫瘤偵測,” 成功大學資訊工程所,
2004.