摘要
针对传统Hopfield神经网络无法对非正交学习模式进行正确回忆的问题,在Hebb学习规则的基础上,提出了一种改进的Hopfield神经网络.通过在学习阶段对连接关系矩阵进行修正,改进后的Hopfield神经网络能够对其学习模式进行正确的回忆,实验证明了其对手写字符有着较好的识别效果.
In light of the problem of recalling non-orthogonal learning models incorrectly in traditional Hopfield neural network, an improved Hopfield neural network is presented on the basis of Hebb Learning Rule. Through modifying the connection relations matrix in learning phase, the improved Hopfield neural network could recall its learning models accurately. Test results indicate that it works efficiently in handwritten recognition.
出处
《天津理工大学学报》
2009年第4期49-52,共4页
Journal of Tianjin University of Technology
