Stochastic Resonance — Suprathreshold SR

Signal & Noise

Signal amplitude A = 0.6 Signal frequency f = 0.8 Hz Noise level σ = 0.5 Threshold θ = 1.0

Population (suprathreshold)

Neuron count N = 8 Threshold spread Δ = 0.3

Performance

SNR (output) = 0.00 dB

Corr(signal,output) = 0.00

Optimal noise ≈ —

Suprathreshold SR: best with population diversity

Physics: Classic stochastic resonance (Benzi 1981) occurs when noise boosts detection of subthreshold signals. Suprathreshold SR (Stocks 2000) extends this to populations of neurons with diverse thresholds — the array collectively encodes information above any single threshold. Even when A > θ, population noise and spread can improve the population signal-to-noise ratio. The output is the sum of binary neuron firings. The optimal noise level maximizes the mutual information between input and output.