Noise driven neuromorphic tuned amplifier

Abstract

Living systems implement and execute an extraordinary plethora of computational tasks. The inherent degree of large scale coordination emerges as a global property, from the intricate sea of microscopic interactions. The brain, with its structural and functional architecture, represents an emblematic example of hierarchic self-organization: elementary units, the neurons, act much like instruments of an orchestra, which combine diverse timbres to create harmonious symphonies. Neurons come indeed in different types, varying in shapes, connections and electrical properties. They all team up to process external stimuli from a number of sources and integrate the information to yield, from neurons to mind, different cognitive faculties. Identifying the coarse grained modules that exert, from bottom to up, pivotal neuronal functions constitutes a goal of paramount importance. On the other hand, the brain and its unraveled secrets could inspire novel biomimetics technologies to adaptively handle complex problems. Here, we investigate the intertwined stochastic dynamics of two populations of excitatory and inhibitory units, arranged in a directed lattice. The endogenous noise seeds a coherent amplification across the chain generating giant oscillations with tunable frequencies, a process that the brain could exploit to enhance, and eventually encode, different signals. The system works as an out-equilibrium thermal device under stationary operating conditions, the associated entropy production rate being analytically quantified. The same scheme could be invoked to design a novel family of manmade detectors capable of reacting to spatially distributed low intensity alerts.