Modeling and Predicting Causal Effects on Complex Networks
Abstract
The ubiquity of complex networks across various domains--including transportation, bio-medicine, economy, military, and social life--has underscored the importance of understanding their intricate dynamics and interdependencies. Recently, leveraging machine learning for complex network analysis has become increasingly popular, aiming to make informed decisions for strategic interventions. Traditional causal inference methods, primarily designed for independent data units, fall short in network contexts due to interconnectedness that leads to spillover effects and treatment entanglement among nodes.This project proposes to explore the theoretical underpinnings and develop computational tools for causal inference in large-scale complex networks. We aim to address three main challenges: 1) the spill-over effect of interventions, which considers the impact of a treatment on interconnected nodes; 2) the effect of entangled interventions, focusing on treatments applied to node clusters rather than isolated nodes; and 3) structured interventions that go beyond binary treatments to include modifications in network topology and dynamic adjustments based on network feedback.Through these research thrusts, we seek to innovate causal inference methodologies that can robustly estimate interventional effects in networked environments, thus supporting high-stakes decision-making in critical areas. In particular, it will significantly enhance Department of Defense (DoD) capabilities by enabling more precise and effective strategies for operations involving communication, surveillance, and cyber-defense networks, thus improving overall mission efficacy and security.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Nov 09, 2024
- Source ID
- N000142412636
Entities
People
- Jundong Li
Organizations
- Office of Naval Research
- United States Navy
- University of Virginia