论文部分内容阅读
对联想记忆神经网络(neuralnetworks NN)的特性进行了分析,基于双向联想存储器BAM原理,对自联想记忆HopfieldNN模型进行了扩展,建造了适合于求解模式识别问题的异联想记忆HopfieldNN模型结构.并针对HopfieldNN记忆容量不足的缺陷,对常规的学习算法进行了改进,建立了基于投影原理的伪逆学习算法和广义逆学习算法,提高了HopfieldNN对样本的记忆存储能力.分析表明:本文的研究解决了HopfieldNN应用于异联想记忆模式识别的关键问题.
Based on the theory of bidirectional associative memory BAM, the self-associative memory HopfieldNN model is extended, and the construction of the orthonormal memory HopfieldNN model which is suitable for solving pattern recognition problems is constructed. HopfieldNN deficiencies of memory capacity, the conventional learning algorithm is improved, based on the projection principle of the pseudo-inverse learning algorithm and generalized inverse learning algorithm to improve the HopfieldNN memory capacity of the sample analysis shows that: This article addresses the HopfieldNN Key Problems in Pattern Recognition of Allied Memory.