论文部分内容阅读
本文通过增加网络的神经元,提出了异联想记忆神经网络模型的快速增强算法,它能存储任意给定的训练模式对集,即对于训练模式对的数目以及它们之间相关性的强弱没有限制.快速增强算法在X-域和Y-域增强的神经元个数分别至多为原先Y-域和X-域的神经元个数.快速增强算法设计出的网络连接权只取值1,0或-1,因而网络易于硬件电路实现和光学实现.计算机实验结果表明,与虚构增强算法相比,快速增强算法大大减少了网络的附加连接权.
In this paper, by increasing the neurons in the network, a fast enhancement algorithm for the heterogeneous associative memory neural network model is proposed, which can store any given pairs of training patterns, that is, the number of pairs of training patterns and the strength of the correlation between them Limitations Rapid enhancement algorithm Increases the number of neurons in the X-domain and Y-domain up to the number of neurons in the original Y-domain and X-domain, respectively. The fast connection-enhancing algorithm has a network connection weight of only 1, 0 or -1, so the network is easy to implement by hardware and optical.Computer experiments show that compared with the fictitious enhancement algorithm, the fast enhancement algorithm greatly reduces the additional connection of the network.