An improved learning algorithm for branch and bound for semi-supervised support vector machines

作者：Zhao, Ying^*; Zhang, Jian Pei; Yang, Jing; Wang, Guan Jun

来源：Acta Electronica Sinica, 2010, 38(2): 449-454.

• Geometric characteristics
• High-precision
• Modified algorithms
• Operation speed
• Semi
• supervised learning
• Semi-supervised
• Statistic theory
• Strong robustness

Branch and bound semi-supervised support vector machines as an exact globally optimization is useful for benchmarking practical semi-supervised support vector machines implementations. An improved learning algorithm for branch and bound for semi-supervised support vector machines is presented, concerning the defects of the branch and bound for semi-supervised support vector machines. The estimations of the node lower bound are redefined, which can reduce time complexity of computing the lower bound on every node. Branching nodes are determined by using the geometric characteristic of the support vector machines, which can improve the operation speed simultaneously. Experimental results show that modified algorithm has high precision and strong robustness.

• 出版日期2010
• 单位哈尔滨工程大学; 中国矿业大学（北京）