Noise and Neuronal Heterogeneity

We consider signal transaction in a simple neuronal model featuring intrinsic noise. The presence of noise limits the precision of neural responses and impacts the quality of neural signal transduction. We assess the signal transduction quality in relation to the level of noise, and show it to be maximized by a non-zero level of noise, analogous to the stochastic resonance effect. The quality enhancement occurs for a finite range of stimuli to a single neuron; we show how to construct networks of neurons that extend the range. The range increases more rapidly with network size when we make use of heterogeneous populations of neurons with a variety of thresholds, rather than homogeneous populations of neurons all with the same threshold. The limited precision of neural responses thus can have a direct effect on the optimal network structure, with diverse functional properties of the constituent neurons supporting an economical information processing strategy that reduces the metabolic costs of handling a broad class of stimuli.