Computational Research on Cognition and Memory Mechanism in Neural Networks with Asymmetric and Symmetric Network Structures

在神经网络中，对空间信息进行具有非线性特征的并行处理将是视觉信息处理的基本原理。对于神经网络中具有非线性特性的并行处理的关键特征，在理论上还没有很好的讨论。研究表明，两层网络级联网络中的非对称非线性函数在运动检测的并行处理中起着至关重要的作用。在此基础上，提出了一种新的并行处理集成处理方法。研究表明，奇偶非线性函数并行处理是一种有效的运动检测方法。用于运动检测的视觉皮层由两层网络组成，称为初级视觉皮质(V1)，随后是中间颞区(MT)。视觉皮层V1区和MT区神经网络结构的基本特征是一个具有非线性通路和线性通路的非对称网络。本文通过对非对称神经网络的分析，讨论了纽约大学Naka教授所阐明的模型神经元。然后将该分析方法应用于视皮层网络。将V1和MT模型网络分解为次不对称网络。通过对非对称网络的优化，推导出运动检测方程。阐明了V1和MT中的奇偶非线性组合非对称网络的集成并行处理是运动检测的基础。从计算的角度来看，V1网络的集成并行处理，其次是MT网络，对运动信息的处理是充分的。将本文提出的集成处理方法应用于文本挖掘问题

In the neural networks, a parallel processing with nonlinear characteristics for the spatial information will be a fundamental principle of the visual information processing. It is not well discussed theoretically to clarify the key features for the parallel processing with nonlinear characteristics in the neural networks. In this research, it is shown that the asymmetrical nonlinear functions in the cascaded network with two network layers, play an crucial role in the parallel processing in the movement detection. Then, an ensemble processing of the parallel processing, is newly developed here. We made clear that the parallel processing with the even and odd nonlinear functions, is an effective in the-movement detection. The visual cortex for the movement detection, consists of two layered networks, called the primary visual cortex (V1), followed by the middle temporal area (MT). The fundamental characteristics of the neural network structure in V1 and MT of the visual cortex, is an asymmetrical network with a nonlinear pathway and a linear pathway The model neurons, clarified by Prof Naka at New York York University are discussed by analyzing the asymmetric neural networks. Then the analysis method was applied to the visual cortex networks. The V1 and MT model networks, are decomposed into sub-asymmetrical networks. By the optimization of the asymmetric networks, the movement detection equations are derived. Then, it was clarified that the ensemble parallel processing with the even-odd nonlinearity combined asymmetric networks in the V1 and MT, are fundamental in the movement detection. It was concluded that the ensemble parallel processing of V1 network, followed by the MT network, process the movement information sufficiently from the view point of the computational aspects. The ensemble processing developed here was applied to the text mining problems