平行分割平面之多類別支撐向量機器

支撐向量機器最初設計的目的是以二元分類為主，然而用傳統的支撐向量機器去做多元分類，在效率跟準確度上都沒有很好的表現，所以至今已發展出許多的方法來做多元分類，而如何處理多元分類是一個很重要的問題。本論文提出將現有支撐向量機器架構下作改進，發展一套新的模式來解決此問題，此模式藉由將訓練資料映射至特徵空間後利用平行的分割平面分隔類別資料。本論文中證明至少存在一個映射函數使得訓練資料經過此映射函數映射至特徵空間後其資料可被平行的分割平面完美分類。本論文提出了類別歸屬初始演算法來解決訓練資料經過映射函數映射至特徵空間後資料的分配位置是否為可辨識的排列方式。經實驗證明本論文所提出的分類方法在準確度上比現有的方法還要好。

關鍵字

平行分割；映射函數；支撐向量機器；多元分類

並列摘要

The support vector machine (SVM) was originally designed for binary classification, has been extended to deal with multi-class classification problem. It is still an open issue although many studies has been proposed to improve the performance of multi-class classification. This paper presents a novel multi-class support vector machine to improve the performance of applying support vector machine on multi-class problem. In the novel scheme parallel hyperplanes are used as the separating hyperplanes to classify patterns. This paper proved that the transformed functions which the patterns are mapped from the data space into the feature space and can be classified by parallel hyperplanes are existed. This paper presents a class degree decision algorithms to predict the class locations in feature space. Experimental results show that our proposed scheme outperform others.

並列關鍵字

Parallel Separating ； Mapping Function ； Support Vector Machine (SVM) ； Multi-Classification

參考文獻

[1] V. N. Vapnik, Statistical Learning Theory,Wiley,Inc., 1 edition, 1998.

[2] Y.Guermeur, "Combining Discriminant Models with New Multi-ClassSVMS," Pattern Analysis & Applications, 2000, Vol. 5, No. 2, pp. 168-179.

[3] E. J. Bredensteiner and K. P. Bennett, “MulticategoryClassification by Support Vector Machines," Computational Optimization, 1999, pp. 53–79.

[5] S. Knerr, L. Personnaz, and G. Dreyfus, “Single-layer learning revisited: A stepwise procedure forbuilding and training a neural network,”Neurocomputing, 1990, pp. 41-50.

[6] U. Kreßel, “Pairwise classification and support vector machines,”Advances in KernelMethods, 1999, pp. 255–268.

國際替代計量

平行分割平面之多類別支撐向量機器

主題瀏覽