10-05-2012, 02:42 PM
How much power does neural signal propagation need
SmartMS02.pdf (Size: 102.06 KB / Downloads: 32)
Introduction
Noise-facilitated signal transduction, or stochastic resonance
(SR) [1], is attracting significant attention (for reviews
see [6, 18, 19]). Here we consider two well known nondynamical
models of noise-facilitated signal transduction from
the point of view of energy dissipation. The first model,
introduced six years ago [7, 10], is a threshold model where a
pulse (or a spike) is generated every time the input parameter
comprised of signal and noise reaches the threshold voltage
value [12]. The second model, described four years ago [2–4],
is a threshold-free model of signal transduction. It is based on
the so-called inhomogeneous Poisson process. In this process
the rateof pulse generation is modulated by the input parameter
in a continuous manner.
Comparison of the models
In the present paper, werestrict our considerations to the case of
small and adiabatically slow signals. We start with Shannon’s
formula for the channel information capacity I to showthat, for
small signals, this measure coincides with the signal-to-noise
ratio (SNR). The channel information capacity characterizes
the rate of information transmission (dimensions: bits/second)
and, for a white spectral distribution of the output noise, can
be written in the form[13]