Parametric Amplification of Signals by Noise in Neurons and Neural Networks

摘要

The positive role of parametric noise in neural systems is shown. In analogy with transistor effect the weak input signals control the intensive response of nonlinear neurons to parametric noise. As a result gigantic amplification of input signals can be obtained. The mechanism of amplification is simple and robust. It can relate both to biological and artificial neural networks. Such amplification was first investigated in a model of overdamped Kramers oscillator [1,2].