A Hybird Approach to Complexity Optimization of Neural Networks

摘要：

In this paper, we propose a hybrid approach to the optimization of complexity of feed-forward neural networks in order to achieve good generalization performances and compact structures. By applying an adaptive regularization method to the network learning, we expect to acquire a good generalization performance of the network. By applying a pruning method to the trained network, we expect to get a compact structure as well as a good generalization performance. We confirm the performance of the proposed method through experiments on a benchmark data set.

作者: Hyunjin Lee Taechang Jee Hyeyoung Park Yillbyung Lee

作者单位: Dept. of Computer Science, University of Yonsei University 134 Shinchon-dong, Seodaemun-gu, Seoul 12 Dept. of R&D Center, LG-EDS Systems 19F, Good Morning Tower, 23-2, Yoido-dong, Youngdeungpo-gu, Seou Lab. for Mathematical Neuroscience, Brain Science Institute, RUCEN 2-1 Hirosawa, Wako, Saitama 351-0

会议类型: 国际会议

会议名称: 8th International Conference on Neural Information Processing(ICONIP 2001)(第八届国际神经信息处理大会)

会议地点: 上海

会议语种:英文

页码: 1493-1498

在线出版日期: 2001-11-14（万方平台首次上网日期，不代表论文的发表时间）