Contagion in Large-Scale Networks: Frameworks for Resource Allocation and Simulation

Contagion in Large-Scale Networks: Frameworks for Resource Allocation and Simulation (Via Zoom)

Abstract: In this talk, I focus on resource allocation in dynamically changing complex environments that undergo contagion in the real world such as peer-to-peer lending networks, viral marketing campaigns, ridesharing, and financial networks. In these systems, a planner needs to allocate resources subject to a budget to cause or prevent contagion. These environments are often very large-scale and dynamically changing and usually have missing links. My work aims to address these two problems.

I present a model for allocations that can be provably and efficiently solved in large-scale networks and is more flexible than existing contagion models. I show the model’s efficiency in several use cases such as physical financial networks, ridesharing networks, and digital financial networks.

Then, I briefly focus on the absence of data in such networks, and the limitations of existing agent-based models. Given the recent successes of Large Language Models (LLMs), I show that collectives of LLMs exhibit several macroscopic characteristics that are similar to human dynamics, such as preferential attachment and long-tailed degree distributions, homophily, and triadic closure. Finally, I suggest several avenues of research based on this paradigm. 

Bio: Marios is a fifth-year Ph.D. candidate in the Computer Science Department at Cornell University,  advised by Prof. Jon Kleinberg. Marios works on algorithms and networks, exploring their roles within large-scale social and information systems, and understanding their wider societal implications. His research has been supported by an Onassis Scholarship, a LinkedIn Ph.D. Fellowship and a Cornell Fellowship, and grants from the A.G. Leventis Foundation and the Gerondelis Foundation.