摘要
为有效剔除学习样本中的病态样本,以提高神经网络的泛化能力,提出一种新的病态样本搜索思想并应用海明距离,给出了基于搜索思想和海明距离的前馈神经网络病态样本剔除方法。该方法不用纳入专家的先验知识、样本形式等其他因素即可完成对病态样本的搜索和剔除,具有较强的适用性。数值模拟实验分析结果表明,该方法是一种科学、有效的方法,借助其能有效地搜索出学习样本中所存在的病态样本,对解决实际问题具有较大的应用价值。
For efficiently eliminating morbid samples and improving the generalization ability of neural networks, we present an approach to eliminate morbid samples in forward neural networks based on the search thought and the Hamming distance, through developing the thought and introducing the distance. The approach can directly carry out searching and eliminating morbid samples, and do not consider prior knowledge, forms of samples, etc. Hence, its applicability is stronger. The results of numerical demonstration analysis show that the approach is scientific, effective and can effectively find out morbid samples of learning samples, it has obvious application value to solve problems of real world.
出处
《吉林大学学报（信息科学版）》
CAS
2009年第5期514-519,共6页
Journal of Jilin University（Information Science Edition）
基金
国家自然科学基金重点资助项目(70732005)
国家自然科学基金资助项目(70471015)
关键词
前馈神经网络
病态样本
剔除
搜索
海明距离
forward neural network
morbid sample
eliminate
search
hamming distance
