Enhanced processing in arrays of optimally tuned nonlinear biomimetic sensors: A coupling-mediated Ringelmann effect and its dynamical mitigation.

Inspired by recent results on self-tunability in the outer hair cells of the mammalian cochlea, we describe an array of magnetic sensors where each individual sensor can self-tune to an optimal operating regime. The self-tuning gives the array its "biomimetic" features. We show that the overall performance of the array can, as expected, be improved by increasing the number of sensors but, however, coupling between sensors reduces the overall performance even though the individual sensors in the system could see an improvement. We quantify the similarity of this phenomenon to the Ringelmann effect that was formulated 103 years ago to account for productivity losses in human and animal groups. We propose a global feedback scheme that can be used to greatly mitigate the performance degradation that would, normally, stem from the Ringelmann effect.

Enhancing the resolution of a sensor via negative correlation: a biologically inspired approach.

We demonstrate that a neuronal system, underpinned by "fire-then-reset" dynamics, can display an enhanced resolution R~T(ob)(-1) where T(ob) is the observation time of the measurement; this occurs when the interspike intervals are negatively correlated and T(ob)<Δ/ε, where ε is a parameter characterizing the level of correlation between interspike intervals and Δ is the average interspike interval. We also show that by introducing negative correlations into the time domain response of a nonlinear dynamical sensor it is possible to replicate this enhanced scaling of the resolution. Thus, we demonstrate the potential for designing a novel class of biomimetic sensors that afford improved signal resolution by functionally utilizing negative correlations.

Neural population coding is optimized by discrete tuning curves.

The sigmoidal tuning curve that maximizes the mutual information for a Poisson neuron, or population of Poisson neurons, is obtained. The optimal tuning curve is found to have a discrete structure that results in a quantization of the input signal. The number of quantization levels undergoes a hierarchy of phase transitions as the length of the coding window is varied. We postulate, using the mammalian auditory system as an example, that the presence of a subpopulation structure within a neural population is consistent with an optimal neural code.