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

在神经网络中，空间信息的并行处理是一个突出的特征。从理论上阐明神经网络并行处理的关键特征并没有得到很好的讨论。研究表明，非对称非线性函数在运动检测的网络并行处理中起着至关重要的作用。视觉信息首先被输入到视网膜神经网络，然后被传递到中继神经元，最后在大脑皮层和中颞叶区的视觉网络中进行处理。在这些网络中，有报道称这种非线性函数可以有效地处理视觉信息。通过对奇偶非线性函数的并行处理，可以有效地进行运动检测。用于运动检测的视觉皮层由两层网络组成，分别是初级视觉皮层（V1）和中颞叶区（MT）。V！和MT模型神经元，通过分析不对称神经网络来讨论。然后，将V1和MT模型网络分解为亚不对称网络。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不对称网络是运动检测的基础。从计算的角度来看，V1网络和MT网络对运动信息的处理是充分的。

In the neural networks, one of the prominent features, is a parallel processing for the spatial information. It is not well discussed theoretically to clarify the key features for the parallel processing in the neural networks. In this research, it is shown that the asymmetrical nonlinear functions play an crucial role in the network parallel processing for the movement detection. The visual information is inputted first to the retinal neural networks, then it is transmitted to the relay neurons and finally is processed in the visual network of the cortex and the middle temporal area in the brain. In these networks, it is reported that this kind of nonlinear functions will process the visual information effectively. We made clear that the parallel processing with the even and odd nonlinear functions, is 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 in V! and MT model neurons, are discussed by analyzing the asymmetric neural networks. Then, 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 even-odd nonlinearity combined asymmetric networks in the V1 and MT, are fundamental in the movement detection. It was concluded that the V1 network, followed by the MT network, process the movement information sufficiently from the view point of the computational aspects.