Affiliations: [a] Post-Doctoral Research Station of China Construction Bank, Beijing, China | [b] Post-Doctoral Research Station of Tsinghua University, Beijing, China | [c] University of Chinese Academy of Sciences, Beijing, China | [d] Peking University, Beijing, China | [e] Langfang Teachers University, Langfang, Hebei, China
Correspondence:
[*]
Corresponding author: Xi Zhao, Post-Doctoral Research Station of China Construction Bank, Beijing, China. Tel.: +86 010 67596805; Fax: +86 010 67596805; E-mail:zhaoxi19850210@163.com
Abstract: Nowadays, learning simulation models with unlabeled data becomes a hot spot in data mining and machine learning fields. Among various kinds of approaches, semi-supervised learning (SSL) has played a more and more important role. In this paper, we proposed a novel method called LapNPSVM for binary classification under SSL scenario. One of the main merits is that our method avoids inversion matrix in objective function compared with laplacian SVM and other twin SVM based approaches which is obviously an big obstacle for large scale application. Experiments on artificial and real world datasets show the generalization and speed effectiveness of our method.
Keywords: Support vector machine, optimization, binary classification
