Targeted interventions and higher-order interactions in complex networks
Abstract
Network interactions play a fundamental role in many social, economic, engineering, and biological systems. It is therefore clear that interventions aimed at increasing resilience, security, or welfare should exploit network information. Modern network systems however are of ever-increasing dimension, maybe time-varying, with agents continuously joining or leaving the system and forming new connections. In these cases, it is impossible for a planner to acquire perfect network information, yet partial or aggregated data can be used to infer a statistical model about the likelihood of agents interactions. Based on this consideration, in this proposal we aim at developing a new class of algorithms for targeted intervention design based on statistical instead of exact network knowledge, with proven guarantees in terms of complexity and performance. Our proposed approach involves two steps. First, by describing statistical network information via suitable notions of graph limits (such as graphons or hypergraphons) we will formally define dynamical systems for networks of infinite size and we will show that, under suitable assumptions, intervention design becomes tractable in such infinite population limit. Second, we will show that interventions computed for the limiting case can be applied (via discretization) to finite networks, yielding approximately optimal performances. We will illustrate the benefits of the proposed procedure for the design of targeted interventions in network contagion processes, focusing on- i)seeding policies to maximize the spread of contagion, with application to product-technology adoption and information diffusion, ii) vaccination policies to minimize the spread of contagion, with application to epidemic containment and network security, and iii) interventions in networks involving higher-order interactions. Our theoretical framework will be validated with case studies involving real network data sets, with broad applicability to several areas of interest to the Department of Defense including cybersecurity, information systems, optimal resource allocation and network resiliency.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Feb 05, 2025
- Source ID
- FA95502410082
Entities
People
- Francesca Parise
Organizations
- Air Force Office of Scientific Research
- Cornell University
- United States Air Force