Common-mode noise suppression via differential readout in paired Rulkov neurons

Background noise limits signal extraction in biological neural systems and neuromorphic circuits, particularly when fluctuations are shared across channels. Here we study an opponent-channel differential readout formed by two parameter-matched Rulkov neurons driven by opposite-polarity inputs and common-mode noise. Across exponential-decay, alpha-function, step, and sinusoidal stimuli, the differential readout improves waveform preservation and reference-aligned signal retention relative to single-neuron and amplitude-matched controls, especially when the shared-noise component is substantial. Robustness tests further show that the advantage decreases under weak noise correlation or large parameter mismatch. Mechanistic analysis reveals that common-mode noise is not completely eliminated in the nonlinear map: it perturbs the common operating state and re-enters the differential channel through state-dependent gain and local slope mismatch. These results identify both the benefit and the nonlinear leakage limit of differential readout in paired Rulkov-neuron dynamics.