在学习过程中引入负反馈的神经网络

在学习过程中引入负反馈的神经网络 THE NEURAL NETWORK WITH NEGATIVE-FEEDBACK IN ITS LEARNING PROCESS 梁明理 Liang Mingli He Hao Wang Yang 1 first-author 何浩 1 汪洋 1 武汉大学物理学系 武汉大学物理学系 Department of Physics,Wuhan University Department of Physics,Wuhan University 本文在 Hopfield 模型的基础上,提出了在学习过程中引入负反馈的神经网络模型。理论分析指出,由新模型构成的神经网络,提高了记忆相关图样的能力和容错能力,记忆容量也远远超过网络中神经元数目的15%。计算机模拟结果表明这些分析的正确性。本文还讨论了新模型与生物神经功能的类似性。 Based on the Hopfield model,the neural network with negative-feedback in its learning process is studied in this paper.Compared with the Hopfield model, the mew model’s ability of recognizing similar patterns is greatly enhanced and both its capacity of allowable errors and memory capacity are enormous The computer simulation conforms those theoretical conclusions.As a result of the negative-feedback,the new model has some similarities to a real neurous system. 神经网络 负反馈 学习过程 记忆 容错能力 neural network negative-feedback learning process memory capacity of allowable errors Q61 1992-02-01 2021-04-01 2