English

应用科学学报

• 论文 • 上一篇    下一篇

基于证据框架的模糊超球面支持向量机超参数优化

金添 周智敏 宋千 常文革   

  1. 国防科学技术大学 电子科学与工程学院 湖南 长沙 410073
  • 收稿日期:2006-04-29 修回日期:2006-11-20 出版日期:2007-05-31 发布日期:2007-05-31

Hyperparameter Optimization of Fuzzy Hypersphere Support Vector Machine Based on Evidence Framework

JIN Tian, ZHOU Zhi-min, SONG Qian and CHANG Wen-ge   

  1. School of Electronic Science and Engineering, National University of Defense Technology, Changsha, 410073, China

  • Received:2006-04-29 Revised:2006-11-20 Online:2007-05-31 Published:2007-05-31

摘要： 模糊超球面支持向量机（FHS-SVM）在处理一类分类问题时比超平面支持向量机泛化能力更强，特别在雷达目标检测中得到了成功应用。FHS-SVM训练时需要预设一些超参数，不同的超参数得到的FHS-SVM性能差异很大。本文首先证明了FHS-SVM训练过程与证据框架第一层贝叶斯推理的等价性，然后在证据框架下提出了FHS-SVM超参数优化迭代方法。基于超宽带合成孔径雷达探雷数据，通过与穷举方法结果的对比检验了迭代优化方法的有效性。

关键词: 证据框架, 模糊超球面支持向量机, 超参数优化, 地雷检测

Abstract: Fuzzy hypersphere support vector machine (FHS-SVM) has stronger generalization capability than hyperplane support vector machine in the one-class classification problem, being successful in radar target detection. Some hyperparameters have to be predefined before the FHS-SVM training, with different hyperparameters leading to significant difference in the FHS-SVM performance. In this paper, equivalence between FHS-SVM training and the level 1 Bayesian inference of the evidence framework is proved. Then, an FHS-SVM hyperparameter optimization iteration method is proposed based on the evidence framework. Using landmine detection data obtained with ultra-wide band synthetic aperture radar, the proposed iteration method is verified by comparing it with an exhaustive search method.

Key words: evidence framework, fuzzy hypersphere support vector machine(FHS-SVM), hyperparameter optimization, landmine detection