Classifier Design via Conic Optimization and Separating Ellipsoids

摘要：

In this paper, new classification systems are introduced. Two different classifiers based on conic optimization are designed. Semidefinite quadratic linear (SQL) optimization problems are solved to obtain optimized ellipsoids that are used in the separation and classification process. The first approach is the proposed upgrade of a 2-class classifier to be able to distinguish between N classes (Voting Classifier). The second approach is the proposed N-ellipsoidal classifier (NEC). Experiments are performed on some data sets from UCI machine learning repository. Results are compared with several well-known classification algorithms, and show that the proposed approaches provide more accurate and less complex classification systems with competitive error rates.

关键词： conic optimization classifier semidefiite ellipsoids

作者: Abdel-Karim S.O. Hassan Mohamed A. El-Gamal Ahmad A. I. Ibrahim

作者单位: Dept. of Eng. Math, and Physics Cairo University Giza, Egypt Dept. of Eng. Math. and Physics Cairo University Giza, Egypt

会议类型: 国际会议

会议名称: 2011 International Conference on Database and Data Mining(ICDDM 2011)(2011年数据库和数据挖掘国际会议)

会议地点: 三亚

会议语种:英文

页码: 200-204

在线出版日期: 2011-03-25（万方平台首次上网日期，不代表论文的发表时间）