Medvedev Georgi S
Department of Mathematics, Drexel University, Philadelphia, PA 19104, USA.
Neural Comput. 2009 Nov;21(11):3057-78. doi: 10.1162/neco.2009.07-08-813.
We consider an integrate-and-fire element subject to randomly perturbed synaptic input and an electrically coupled ensemble of such elements. The latter is interpreted as either a model of electrically coupled population of neurons or a multicompartment model of a dendrite. Random fluctuations blur the input signal and cause false responses in the system dynamics. For instance, under the influence of noise, the system may respond with an action potential to a subthreshold stimulus. We show that the responses of the elements within the network are more reliable than the responses of the same elements in isolation. Specifically, we show that the variances of the stochastic processes generated by the coupled model can be made arbitrarily small (i.e., the network responses can be made arbitrarily accurate) by increasing the number of elements in the network and the strength of electrical coupling. Our results suggest that the organization of cells in electrically coupled groups on the network level, or the dendritic morphology on the cellular level, may be involved in the filtering noise and therefore may play an important role in the information processing mechanisms operating on the network or cellular level respectively.
我们考虑一个受随机扰动突触输入影响的积分发放元件以及由这类元件构成的电耦合集合。后者可被解释为电耦合神经元群体的模型或树突的多房室模型。随机涨落会模糊输入信号并在系统动力学中引发错误响应。例如，在噪声影响下，系统可能会对阈下刺激产生动作电位响应。我们表明，网络内元件的响应比单个元件单独的响应更可靠。具体而言，我们表明，通过增加网络中元件的数量和电耦合强度，耦合模型所产生的随机过程的方差可以被任意减小（即网络响应可以被任意精确）。我们的结果表明，网络层面上电耦合组中细胞的组织方式，或细胞层面上的树突形态可能参与了噪声过滤，因此可能分别在网络或细胞层面上运行的信息处理机制中发挥重要作用。
