Controlling Epidemics on Networks and Explaining Spatio-Temporal Clusters of Epidemic Spread
Abstract:
As the COVID-19 outbreak has shown, epidemic outbreaks can lead not only to a public health crisis --- killing millions of people and straining the public health workforce and resources --- but also to potential economic and mental health crises. Many complex factors drive the spread of epidemics. Therefore, understanding the characteristics of an epidemic spread and developing interventions to control it are fundamental public health problems. We study two such problems in this dissertation. The first problem, which we refer to as \textit{Epidemic Control}, involves designing interventions (e.g., vaccinations), to mitigate the outbreak. This is a challenging stochastic optimization problem in the context of the SIR model of an epidemic on a network. We have developed efficient approximation algorithms for this problem using techniques from stochastic optimization. These algorithms provide empirical guarantees for the solution quality in graphs of moderate size. We propose to extend these algorithms so that they scale to much larger networks and capture more complex epidemic models. The second problem, which we refer to as \textit{Explaining Spatio-temporal clusters of epidemic spread}, involves identifying spatio-temporal patterns in epidemic spread by modeling it as a cluster explanation problem. We present approximation algorithms and practical heuristics for this problem using techniques from submodular optimization. We propose to (i) make these methods more efficient in practice and (ii) utilize these methods to explain spatio-temporal clusters in the data collected from the spread of the COVID-19 pandemic.
Committee:
- Madhav Marathe, Committee Chair (Biocomplexity Institute, CS/SEAS/UVA)
- Anil Vullikanti, Advisor (Biocomplexity Institute, CS/SEAS/UVA)
- Abhijin Adiga (NSSAC/Biocomplexity/UVA)
- Jundong Li (CS, ECE/SEAS, DSI/UVA)
- Michael Porter (ESE/SEAS/UVA)
- B. Aditya Prakash (CS, Georgia Institute of Technology)