The brain network exhibits a remarkable hierarchy of timescales: early sensory areas respond rapidly to stimuli, while higher-order association areas integrate information over extended periods. This hierarchical temporal organization, despite dense inter-areal connectivity and widespread feedback loops, raises a fundamental question: how does the cortex maintain such localized timescales without interference from upstream dynamics? In this talk, I will present a large-scale dynamical model of the brain network that reveals a novel regime of interference-free propagation. In this regime, the cortical network balances synaptic drive such that temporally fluctuating components from upstream inputs are selectively balanced, allowing each area to preserve its intrinsic timescale while reliably transmitting mean signals downstream. I will also discuss how this mechanism generalizes to spiking and other nonlinear networks capable of performing distributed decision-making. The findings offer a unifying principle that supports both stable hierarchical dynamics and efficient cortical communication, shedding light on how the brain flexibly orchestrates complex functions.

李松挺, 上海交通大学自然科学研究院、数学科学学院教授, 教育部长江学者。研究方向为应用数学与计算神经科学, 研究成果发表在CPAM, PNAS, PLoS Computational Biology, NeurIPS等国际期刊和会议上。现担任中国神经科学学会计算神经科学分会副主任和CSIAM, IEEE子刊等多个国际期刊编委, 并获得全国高校青年教师教学竞赛理科组一等奖、上海市青教赛特等奖、宝钢优秀教师奖、上海市五一劳动奖章、上海交大第四届十大科技进展等荣誉。