基于SVC 和SVR 约束组合的迁移学习分类算法

根据迁移学习思想,针对分类问题,以支持向量机(SVM)模型为基础提出一种新的迁移学习分类算法CCTSVM.该方法以邻域间的分类超平面为纽带实现源域对目标域的迁移学习.具体地,以支持向量分类的约束条件完成对目标域数据的学习,获取分类超平面参数,再以支持向量回归的约束条件有效利用源域数据矫正目标域超平面参数,并在上述组合约束的共同作用下实现邻域间迁移,提高分类器性能.在人工和真实数据集上的实验表明,所提出算法具有良好的迁移能力和优越的分类性能.

Transfer classification learning based on combination of both SVC and SVR’s constraints

Transfer learning focuses on solving learning tasks in the target domain by leveraging the useful information in the target domain. Therefore, a novel transfer classification learning algorithm called CC-TSVM is proposed, which is based on SVM learning framework. CC-TSVM firstly adopts SVC to get the hyperplane for the source domain and then corrects the obtained hyperplane as the hyperplane for the target domain by using SVR with the constraints for the labeled data in the target domain. The experimental results on artificial and real datasets show that the proposed algorithm has high classification accuracy and can well leverage the useful information in the source domain.