摘要
提出了一种新的用于双向联想记忆的学习算法,该算法利用了输入向量各元素之间的关联信息,在联想的过程中,动态地调整权值矩阵,增强了网络适应能力,利用了更多的已知信息,从而提高了网络的性能.
A novel learning algorithm of bidirectional associative memory is proposed in this paper, which makes use of more correlation information than others. The information utilized includes the correlation among the elements of not only the sample patterns, but also the input pattern. The network could adjust the weight matrix dynamically according to the input pattern. All of this improve the adaptability and the performance of the network.
出处
《小型微型计算机系统》
CSCD
北大核心
2005年第6期976-978,共3页
Journal of Chinese Computer Systems
基金
国家自然科学基金(10402003)资助
北京理工大学基础研究基金(BIT-UBF-200301F10)资助
关键词
神经网络
双向联想记忆
学习算法
neural network
bidirectional associative memory
learning algorithm
