摘要
本文基于遗传算法和级联相关算法的组合,提出了遗传辅助级联相关学习算法。该算法能够根据问题的需要自动设计网络结构,同时能有效地消除竞争协议问题。实验结果表明,与级联相关算法相比,它具有结构紧凑,泛化能力好等优点。
Based on genetic technique and cascade correlation algorithm,a new constructive algorithm for connectionistlearning,genetic aided cascade correlation algorithm,is proposed It can automatically grow the architecture ofne ural network to give a suitable n etwork size for a spoc ific problem and is free of the competing conventionsproblem. Experimental results show that the network obtained with our technique is of small size and has ex-cellent performance compared with the normal cascade correlation learning algorithm.(
出处
《核电子学与探测技术》
CAS
CSCD
北大核心
1996年第3期174-176,224,共4页
Nuclear Electronics & Detection Technology
基金
国家自然科学基金
关键词
遗传算法
级联相关
神经网络
Geneticalgorithm Cascade correlation Neural network)