Algorithm of reducing support vector set

作者：Zhao Hui^*; Rong Li Li

来源：Journal of Dalian University of Technology, 2006, 46(5): 735-740.

• Burges method
• Classification speed
• Downs method
• Large-scale training set
• Reducing support vectors
• Support vector machines
• Weights

The standard algorithms for training support vector machines generally produce solutions with a larger number of support vectors than are strictly necessary. Unnecessary support vectors have negative effects on support vector machines'classification speed and practical application. A new method is presented in order to reduce support vector set. Furthermore, an algorithm is proposed which recognizes and eliminates unnecessary support vectors from the original support vector set and computes the new reduced support vector set and its weights. The new method is especially suitable to the case of large-scale training set and large number of support vectors. The experimental results indicate that the new method can remarkably reduce the number of support vectors and increase the speed of classification in the condition that the correct rate does not decline.

• 出版日期2006
• 单位大连大学