论文部分内容阅读
本文提出一种双向联想记忆神经网络的按‘位’加权编码策略，并给出了求取权值的速推算法．它将Ｋｏｓｋｏ双向联想记忆神经网络按海明距离进行模式匹配的原则，修正为按加权海明距离进行模式匹配，从而可以使得对不满足连续性的所谓“病态结构”的一类样本模式集，同样具有良好的联想能力．对二值图象模式存贮、联想的计算机模拟实验表明，此方法具有优良的性能和实用价值．
In this paper, we propose a bit-wise weighted coding strategy for bi-directional associative memory neural networks and give a fast push algorithm for weighting. It modifies the Kosko bidirectional associative memory neural network according to the principle of pattern matching by Hamming distance and modifies the pattern matching by the weighted Hamming distance so as to make it possible to classify a set of sample patterns called “pathological structures” that do not satisfy the continuity, The same has a good ability to think. The binary image mode storage, Lenovo’s computer simulation experiments show that this method has excellent performance and practical value.