摘要
以附加神经元引入附加背景的方式获得将线性离散比极神经元的神经网络在单通道光学矢量-矩阵乘法器内实现的方法,给出了相应的光学系统的修正和非负光学模板的编码形式.以双极神经元的双向联想存储器为例进行了计算机和光电实验模拟.
A general approach to implement bipolar neural network in a single channel vector-matrix multiplier is proposed. Bipolar performances are achieved by using unipolar data and introducing an additional neuron into the network. Minor modifications of the optical system and the formation of the enlarged nonnegative memory matrix are presented. Results of computer simulation and optical demonstration in case of bipolar bidirectional associative memory are also provided.
出处
《光学学报》
EI
CAS
CSCD
北大核心
1993年第4期335-339,共5页
Acta Optica Sinica
基金
国家自然科学基金资助
关键词
神经网络
存储器
光学系统
模拟
optical implementation of neural network, bipolar neuron, bidirectional associative memory.