Criticality, computation and avalanches in neural dynamics

Abstract

The hypothesis of computation at the edge of chaos, which claims systems in critical states should exhibit optimal computational capabilities, suffered several criticisms when was proposed about twenty years ago. However recent studies from neuroscience and neural networks have given new support to that proposal.This project describes several research proposals on the characterization of the information processing capacity of theoretical models conceived to explain the empirically observed critical behavior of samples of cortical tissue both in vitro and in vivo. Specifically, multielectrode arrays spatially disposed in that systems reveal power-law activity patterns for both the size and the duration of such neuronal avalanches.The predominant theoretical framework in the interpretation of that results is self-organized criticality (SOC). Notwithstanding, there are stochastic models of spiking neurons that apparently exhibit the critical behavior of the neuronal avalanches, despite lacking any critical dynamics. This project proposes the investigation of the computational capabilities of SOC models inspired by neuronal avalanches [Nature Phys. 3, 857 (2007), Nature Phys. 6, 801 (2010)], of a minimal model that jointly presents statistical and dynamical criticality [Phys. Rev. Lett. 102, 258102 (2009)] and the investigation of the computational capabilities of noncritical stochastic models of spiking neurons, like [PLoS Comput. Biol. 6:e1000846 (2010)]. Among other objectives, we hope these investigations can indicate whether the notion of computation at the edge of chaos is relevant to the understanding of the critical behavior of neuronal avalanches.Besides that, as long as criticisms on the relevance of the SOC formalism to the understanding of neuronal avalanches have been arising, we propose an investigation on the role of SOC-like arguments in the interpretation of the model proposed in [Nat. Phys. 6, 801 (2010)] to jointly describe neuronal avalanches and the up-down alternance in some regions of the cerebral cortex and also the study of the genericity of stochastic spiking neuron models in generating the critical phenomenology of avalanches. (AU)