Improvements on Twin Support Vector Machines

Yuan-Hai Shao, Chun-Hua Zhang, Xiao-Bo Wang, Nai-Yang Deng
2011 IEEE Transactions on Neural Networks  
Considering the multiple instance learning(MIL) in classification problem, a novel multiple instance twin support vector machines(MI-TWSVM) method is proposed. For linear classification, unlike other maximum margin SVM-based MIL methods, the proposed approach leads to two non-parallel hyperplanes. The non-linear classification via kernels is also studied. Preliminary experimental results on public datasets indicate that our MIL method is competitive with the previous MIL methods.
doi:10.1109/tnn.2011.2130540 pmid:21550880 fatcat:go72cy3le5e7vn5uknsm53ncyy