Noise in the brain: a physical network model

In the brain as in any other open physical systems, noise is inevitable. We present an explicit model of an active physical system that is borrowed from laser physics and allows us to establish the properties of the fluctuating forces that cause noise in the system. It is shown how the cooperation of the individual parts of a system (atoms or neurons) can considerably reduce the noise level. In particular we determine the correlation function between the individual parts. The basic equations can be transformed in such a way that a close analogy with typical equations of neural nets are obtained. In particular, the nonlinear properties of neurons described by the sigmoid function are well captured. Propagation of excitation in axons and dendrites is represented by a linear equation, where we consider both a bandwidth filter and more or less free propagation. In the latter case, a close analogy with an equation for neural activity in the sense of Nunez and established by Jirsa and Haken is pointed out.