神经网络的动力学行为与信息处理

DYNAMICAL BEHAVIOR AND INFORMATION PROCESSING IN A NEURONAL NETWORK

从一个神经元模型出发，研究了由１０００个神经元组成的网络中神经元动作电位发放的动力学特性。由神经元动作电位发放的动态时间—空间图形和时间间隔序列，我们发现在周期性外刺激下网络动力学行为表现出复杂的时空结构。通过调节参数Ｔ１，Ｔ２和Ｊ，我们讨论了网络的不同激发模式和网络中信息传递特性，发现周期性外刺激明显地对神经网络的状态具有调制作用，网络的同步激发模式对应于较简单的时空图形，是一类较为简单的信息处理过程。

Dynamical behavior of the interspike time interval sequences and the structure of spatiotemporal patterns of firings of neurons in a coupled neuronal network are investigated.We also study the information processing in the network. We construct a uniform globally coupled network consisting of one thousand neurons. Under the periodic stimuli,there exist complicated spatiotemporal patterns.The pattern is modulated by the impulse-like synaptic current which is globally coupling interaction between neurons. The firings of neurons are phase-locked to this periodic current. With various parameters of the external stimulus, T1 、T2 ,and coupling strength J, we discuss different firing patterns and the properties of information processing in the network. External stimulus parameters T1 and T2 have more important impacts on states of neurons than the coupling strength. The spatiotemporal pattern is chaotic or most of neurons fire chaotically when T1 <5. As T1 increases,the orderly part of the pattern is increased or the chaotic firing decreases,with more and more neurons being synchronized for 5< T1 <15. When 15< T1 <25,the spatiotemporal pattern is regular or orderly,and almost all neurons show a synchronous firing . As T1 furthermore increases,though there exist chaotic firing areas between different clusters of the neurons which fire with the same period, most neurons are still synchronized together. The influence of T2 on states of neurons is dependent on the valus of T1 . For a fixed T1 and J, the number of nonfiring neurons decreases and the averaged firing rate increases as T 2 increases.The system has a transition from the disorderly to the orderly state is due to the crease of the summation of the spatial and temperal effects of the membrane potential which behaves a strong nonlinear feature.The synaptic current which is the force mediating the interactions between neurons is periodic reacting to the external stimuli though it is sometimes noisy. It is this strong periodic reaction that make the whole system keep the synchronizd behavior across the network.We beweve the synchronized firings of neurons correspond to relatively simple information processing.