Speaker: 纪鹏 (复旦大学类脑智能科学与技术研究院)
Time: 2026-03-05 10:00-11:00
Location: 1610, SIMIS
Abstract:
Perturbations propagating through networked systems underlie diverse collective phenomena, from epidemic outbreaks to cascading failures. Understanding how network topology and nonlinear dynamics jointly shape propagation remains a fundamental challenge. In previous work, we showed that basic network motifs (e.g., triangles) fundamentally alter perturbation propagation, overturning classical scaling regimes (distance-limited vs. degree-limited) that ignore local structures. However, these findings were model-dependent. In ongoing work, we develop a general framework using non-equilibrium response functions that systematically capture how perturbations propagate through arbitrary networks, incorporating topology, nodal dynamics, and interactions. These functions reveal time-dependent effective connectivity and quantify direct versus indirect pathway contributions. Validated on zebrafish whole-brain recordings, our reconstructed response functions predict neural dynamics more accurately than anatomy-based models. Together, these studies establish a unified pathway from identifying key local structures to building a cross-system theory of perturbation propagation, with potential implications for improving AI architectures by inspiring more robust information propagation mechanisms and enhancing interpretability in neuromorphic computing.
About Speaker:
纪鹏，理学博士，复旦大学类脑智能科学与技术研究院研究员、博士生导师。2015年获德国柏林洪堡大学理论物理博士学位，之后在德国波茨坦气候影响研究所工作，2017年加入复旦大学，先后担任青年研究员、研究员。曾入选上海市浦江人才、上海高校特聘教授（东方学者）及东方学者跟踪计划。目前从事的研究涉及复杂系统、数据驱动分析、机器学习算法等交叉学科研究。以第一作者或通讯作者在 Nature Physics、Nature Communications、Physics Reports、Physical Review Letters、Physics of Life Reviews 等期刊发表多篇论文。担任 Physics of Life Reviews、Chaos、Data Analytics and Topology 等学术期刊编辑。
English 
简体中文