FisMat2017 - Submission - View

Abstract's title: Noise driven neuromorphic tuned amplifier
Submitting author: Duccio Fanelli
Affiliation: Dipartimento di Fisica e Astronomia, University of Florence
Affiliation Address: via Sansone 1 50019 Sesto Fiorentino, Italia
Country: Italy
Oral presentation/Poster (Author's request): Oral presentation
Other authors and affiliations: Francesco Ginelli (University of Aberdeen), Roberto Livi (University of Florence), Niccolò Zagli (University of Florence), Clement Zankoc (University of Florence and University of Aberdeen)

Living systems execute an extraordinary plethora of complex functions, that result

from the intertwined interactions among key microscopic actors. Positive and

negative feedbacks appear to orchestrate the necessary degree of macroscopic

coordination, by propagating information to distant sites while supporting the

processing steps that underly categorization and decision making. Excitatory and

inhibitory circuits play, in this respect, a role of paramount importance. As an

example, networks of excitatory and inhibitory neurons constitute the primary

computational units in the brain cortex and can adjust to dierent computational

modalities, as triggered by distinct external stimuli. Genetic regulation also relies on

sophisticated inhibitory and excitatory loops.


Working in this context, we shall here discuss a minimal model for a discrete

collections of agents in mutual interaction via excitatory and inhibitory loops, bearing

universality traits in light of its inherent simplicity. Endogenous-noise stemming from

finite size corrections induces quasi-cyclic dynamics that display unusual long range

correlations, persisting over arbitrary large network structures. When the excitatory

and inhibitory species are distributed on a directed network, the internal noise seeds

giant quasi-cycles, with tunable frequency and amplitude. The system spontaneously

behaves as an effective, stochastic driven pacemaker, a non trivial self-organized

dynamics that holds general interest, for its fundamental to applied implications. The

phenomenon is characterized analytically. The theory prediction are corroborated by

direct stochastic simulations.