An improvement of the design method of cellular neural networks based on generalized eigenvalue minimization
An improvement of the design method of cellular neural networks based on generalized eigenvalue minimization
复制标题
基于广义特征值最小化的细胞神经网络设计方法的改进
DOI：
10.1109/tcsi.2003.819827
复制
发表时间：
2003
影响因子：
5.1
通讯作者：
T. Nishi
中科院分区：
文献类型：
--
作者：
Ryoma Bise;Norikazu Takahashi;T. Nishi
研究了通过具有二进制输出的细胞神经网络（CNN）实现联想记忆。针对这个问题，最近提出了一种基于广义特征值最小化（GEVM）的CNN设计方法。在本文中，将提出一种基于 GEVM 方法的新 CNN 设计方法。我们首先给出一些与记忆向量吸引盆相关的分析结果。然后，我们将这些分析结果与基于 GEVM 的方法相结合，得出设计方法。我们最终通过计算机模拟表明，所提出的方法比原始的基于 GEVM 的方法可以获得更高的召回概率。
Realization of associative memories by cellular neural networks (CNNs) with binary output is studied. Concerning this problem, a CNN design method based upon generalized eigenvalue minimization (GEVM) has recently been proposed. In this brief, a new CNN design method which is based on the GEVM-based method will be presented. We first give some analytical results related to the basin of attraction of a memory vector. We then derive the design method by combining these analytical results and the GEVM-based method. We finally show through computer simulations that the proposed method can achieve higher recall probability than the original GEVM-based method.