Cascading failures in power grids and their containment
What if cascading failures in power grids and online information diffusion in the blogosphere are governed by similar viral diffusion models ?
In this line of work, we model cascading failures in power grids as an information diffusion process, like rumor spread in social networks. We use a Markovian, local stochastic diffusion model that combines viral diffusion principles with physics-based concepts. This is achieved by correlating diffusion weights (contagion probabilities between grid transmission lines) with a hyperparametric Information Cascades model. We show that such a diffusion model can be learned from observed failure data, allowing for accurate modelling and prediction of failure propagation. We then investigate the diffusion containment problem, by a hyperparametric influence minimization model. This model integrates the hyperparametric diffusion framework into the classical influence minimization paradigm, enabling practical and fine-grained control over diffusion dynamics through feature interventions on nodes. The goal is to minimize the expected size of potential diffusions from initial failure points, by optimizing node feature interventions. We analyze the challenges and properties of hyperparametric influence minimization, and we describe several greedy algorithms for this problem. This work was presented in two recent research papers (ACM KDD 2024 and 2025).
Speaker’s profile
Dr. Cautis obtained his PhD in Computer Science from University of Paris-Sud 11 in 2007, and the MSc degree in Computer Science from Ecole polytechnique, in 2004. He received his Habilitation to Supervise Research (HdR) in 2012, from Université Pierre et Marie Curie. Since J une 2025, he is an Associate Professor at Singapore Institute of Technology (SIT) in the Infocomm Technology Cluster (ICT). Before that, he was a Professor at University of Paris-Saclay. He is also a Visiting Professor at the University of HongKong, since 2018. His research is in the areas of data mining and data management, with a specific focus on graphs. He publishes regularly in top-tier conferences (SIGMOD, KDD, VLDB, IJCAI, ECML-PKDD, ICDM, etc) and journals (TODS, TKDD, TKDE, Springer DAMI, etc) on AI and databases. Since 2007, he has supervised eleven PhD students and several postdoctoral researchers. Between 2015 and 2017, he was also a visiting researcher in an industrial lab, Huawei Noah’s Ark Lab, in Hong Kong. Since November 2021 he is a Lead Principal Investigator in DesCartes, a CNRS-managed Singapore-France joint research program (funded by the NRF-CREATE) on intelligent modelling for decisionmaking in critical urban systems.
For more information about the ESD Seminar, please email esd_invite@sutd.edu.sg