Neural criticality has emerged as a unified framework that reconciles diverse multiscale neural dynamics such as the irregular firing of individual neurons, sparse synchrony in neural populations, and the emergence of scale-free avalanches. However, the functional role of neural criticality remains ambiguous. In particular, how highly sensitive and variable critical neural dynamics coexist with reliable neural representation is an unanswered question. In this study, we investigate the neural dynamics and representations in response to external signals in excitation–inhibition (E-I) balanced networks. Our findings reveal that, in contrast with the traditional critical branching model, the critical state of the balanced network simultaneously achieves maximal response sensitivity, maximal response reliability, and optimal representation of external signals. Moreover, we unveil a functional role of neural avalanches, which are not only a hallmark of criticality but also direct contributors to neural representation. By generalizing a semi-analytical mean-field theory, we demonstrate that heterogeneity in inhibitory connections is a mechanism underlying the heterogeneity of steady-state firing rates in the neural population, enabling the presence of signal-reliable neurons in reliable avalanches. Our study addresses a longstanding challenge concerning the functional significance of neural criticality -- the intricate coexistence of reliability and variability.

Reference: 

Zhuda Yang, Junhao Liang and Changsong Zhou, Critical neural avalanches reconcile response reliability with sensitivity for optimal neural representation, Physical Review Letters (in press, Editor’s suggestion). 

报告人简介

周昌松，物理学博士，香港浸会大学物理系物理和复杂系统讲座教授，现任浸会大学非线性研究中心主任，计算及理论研究所所长，生命科学影像中心主任，曾任物理系系主任（2021-24）。1992年获南开大学物理学士，1997年获南开大学物理博士，1997-2007年在新加坡、香港、德国等地从事访问研究，是洪堡基金获得者。2007年加入香港浸会大学物理系，2011年获香港浸会大学“杰出青年研究者校长奖”，2021年获“杰出研究表现校长奖”，及香港研究资助局“高级研究学者奖” (2023）。周昌松博士致力于复杂系统动力学基础研究及其应用，特别是网络的复杂联结结构与体系的动态行为的关系和相互作用。近几年与国际国内系统和认知神经科学家合作，把这些理论进展应用到大脑的复杂联结结构和活动以及认知功能及障碍的分析和建模等方面研究中。周昌松博士对生物神经网络复杂结构、动力学及其低成本高效益如何启发类脑智能具有浓厚的兴趣。在国际交叉学术刊物PNAS，Nature Communications, Physical Review Letters，Physics Reports，National Science Review，J Neuroscience， NeuroImage，Cerebral Cortex，PLoS Computational Biology等发表论文170余篇 (Google Scholar引用19600余次，H-因子为54)。任Scientific Reports 编委，PLoS One，Cognitive Neurodynamics学术编辑，及多种国际期刊常任审稿人。

邀请人

周海军 研究员