动态神经元网络模型的复杂性问题

在动态神经元网络模型中，当神经元总数仅为３时就观察到了非周期振荡。运用Ｌｅｍｐｅｌ和Ｚｉｖ提出的复杂性度量对这种现象进行了分析，结果表明对于其中一个神经元所发出的脉冲序列来说，至少直到１０００个脉冲为止还不能发现任何的周期性，并且其复杂性可以和由ｌｏｇｉｓｔｉｃ映射所产生的时间序列当其参数落在混沌区中时所具有的复杂性相比拟．这些结果也表明这种方法是所观察的时间范围内区分长周期振荡和非周期活动的好方法。结果还提示神经生理实验记录中所谓的噪声，其中有些可能是来源于生物神经元本身的非线性性质。

COMPLEXITY OF A "SIMPLE" REALISTIC NEURAL NETWORK MODEL

Irregular oscillation was observed in observed in our realistic neural network model composed of onlythree neurons under constant external stimuli.Such phenomenon was analyzed using some complexity measure developed by Lempel and Ziv.The results showed that at least for animpulse train of a neuron,up to 1000 impulses showed no trace for any periodicity,and its complexity could be compared with a temporal series series from a logistic mapping while its parameter was within chaotic areas.The results also showed that this measure was a good wayto distinguish long period oscillation from a nonperiodic oscillation within the time domain wecould observe.The above results hint that some of the so called noise in neurophysiological recording might be due to the nonlinear nature of living neurons.