摘要
本文首先对饱和模式线性神经网络模型(LSSM)进行分析,指出其实际上是一种沿内部和边界面依次搜索局部极小值点进行存储的联想记忆模型,进而证明了基于约束区域的神经网络与LSSM具有相同的联想记忆特性,它由普通的常微分方程描述,更利于模拟和硬件实现,计算机模拟说明了结论的正确性。
This paper first analyses LSSM, and prove it to be a kind of associate memory neural network model which searchers for local minimal points along its interior and border. The neural network model based on the constraint domain [6] is proved to have the same associative memory properties as LSSM. The proposed network is described by a ordinary differential equation, and it can be simulated and implemented easily. The simulated results by computer show our conclusions are reasonable.
出处
《模式识别与人工智能》
EI
CSCD
北大核心
2000年第1期28-31,共4页
Pattern Recognition and Artificial Intelligence
关键词
LSSM
约束区域
二次规划问题
神经网络
LSSM, Constraint Domain, Quadratic Programming Problems, Neural Network
