Growth of stochastic resonance in neuronal ensembles with the input signal intensity.

The presence of noise can improve the response of certain nonlinear systems to input signals through the effects of stochastic resonance (SR). The optimal noise intensity for SR is proportional to the signal frequency if the signal is periodic, but proportional to the signal intensity if the signal is aperiodic. Here, we demonstrate using linear response theory that the optimal noise intensity for SR is necessarily dependent on the signal intensity even if the signal is periodic. We also demonstrate that the SR curves grow according to the signal intensity from a bell-shaped curve to a plateau, resulting in the emergence of SR without tuning. In particular, we present a theoretical analysis indicating that the SR peak shifts with the signal intensity due to the scaling of the stationary neuronal firings. The growth of SR may serve as a useful design principle for many noise-exploiting applications.