计算机与现代化 ›› 2009, Vol. 1 ›› Issue (7): 1-4,43.doi: 10.3969/j.issn.1006-2475.2009.07.001

• 人工智能 •    下一篇

基于前馈神经网络的增量学习研究

刘建军;胡卫东;郁文贤   

  1. 国防科技大学电子科学与工程学院ATR重点实验室,湖南 长沙 410073
  • 收稿日期:2008-07-01 修回日期:1900-01-01 出版日期:2009-07-10 发布日期:2009-07-10

A Survey of Research Work on Incremental Learning Based on Feedforward Neural Networks

LIU Jian-jun; HU Wei-dong; YU Wen-xian   

  1. State Lab of Automatic Target Recognition, National University of Defense Technology, Changsha 410073, China
  • Received:2008-07-01 Revised:1900-01-01 Online:2009-07-10 Published:2009-07-10

摘要: 增量学习是一种在巩固原有学习成果和不需要用到原有数据的情况下快速有效地获取新知识的学习模式。本文阐述了基于前馈神经网络的增量学习原理,在此基础上对主要的增量学习算法进行了详细的介绍和分析,最后对增量学习研究进行了总结和展望。

关键词: 模式识别, 神经网络, 增量学习

Abstract: Incremental learning mode is meaningful to efficiently learn new information without forgetting previously acquired knowledge and without requiring access to the original database. The incremental learning principles of feedforward neural networks and feedforward neural network ensembles are expounded. The main incremental learning algorithms based on the principles are introduced and analyzed in detail. Finally, the future direction of incremental learning is depicted.

Key words: pattern recognition, neural networks, incremental learning

中图分类号: