基于覆盖思想的构造型神经网络的增量学习

增量学习是在原有学习成果的基础上,对新信息进行学习,以获取新知识的过程,它要求尽量保持原有的学习成果.文章先简述了基于覆盖的构造型神经网络,然后在此基础上提出了一种快速增量学习算法.该算法在原有网络的分类能力基础上,通过对新样本的快速增量学习,进一步提高网络的分类能力.实验结果表明该算法是有效的.

Incremental Learning of Cover Based Constructive Neural Networks

Incremental learning,which is based on primary learning achievement,learns the new information to gain new knowledge.It requires to keep primary learning achievement.the cover algorithm is introduced firstly,and then a new incremental learning algorithm which can improve the ability of classification by fast training on new samples is investigated.The experiments show the algorithm is effective.