A Hybrid Mechanistic Agent-Based Modeling and Machine Learning Approach for Epidemic Mitigation Design
Project Description
Simulating complex dynamics over multiple generations and large populations is computationally demanding. A single ABM run can take several hours, and full sensitivity analyses often require thousands to millions of such simulations. This computational demand makes brute-force parameter sweeps impractical. To address this challenge, this project will integrate heteroskedastic Gaussian processes (hetGP) as a surrogate modeling technique [1]. Note that hetGP has been shown to reduce computation times by over 90% in high-dimensional simulations, decreasing training time from 339 to 29 seconds in a six-dimensional parameter space.
By approximating ABM outputs using a smaller number of strategically selected simulations, hetGP enables efficient exploration of large parameter spaces with minimal loss of accuracy. Once the surrogate model is built, machine learning tools such as Random Forests and Neural Networks will be used to analyze the relationship between input parameters and resistance outcomes. These tools not only rank variables by their relative importance but also help discover non-linear interactions and identify key factors in the success of different intervention strategies, such as targeted testing-for-isolation and quarantine strategies
Start Date
As soon as possible.
Postdoc Qualifications
The ideal candidate will have strong experience in computational modeling, AI, and/or control, preferably applied to epidemic systems, excitement to develop and innovate new models and methods, excellent communication and collaborative skills, and independent drive. The candidate should also have a strong publication record for their career stage and field of expertise.
Co-advisors
Philip E. Paré, philpare@purdue.edu, ECE, https://sites.google.com/view/philpare
Qixin He, heqixin@purdue.edu, BIO, https://www.qixinhe.net
Bibliography
[1] M. Binois and R. B. Gramacy, “hetGP: Heteroskedastic Gaussian Process Modeling and Sequential Design in R,” Journal of Statistical Software, vol. 98, pp. 1–44, 2021.
[2] P. E. Paré, C. L. Beck, and T. Başar, “Modeling, Estimation, and Analysis of Epidemics over Networks: An Overview,” Annual Reviews in Control: Special Issue on Systems & Control Research Efforts Against COVID-19 and Future Pandemics, 2020.
[3] A. R. Hota, J. Godbole, and P. E. Paré, “A Closed-Loop Framework for Inference, Prediction, and Control of SIR Epidemics on Networks,” IEEE Transactions on Network Science and Engineering, 2021.
[4] Q. He, S. Pilosof, K.E. Tiedje, S. Ruybal-Pesántez, Y. Artzy-Randrup, E.B. Baskerville, K.P. Day, M. Pascual. “Networks of genetic similarity reveal non-neutral processes shape strain structure in Plasmodium falciparum,” Nature Communications, 9(1):1817, 2018.
[5] Q. He, J.K. Chaillet, F. Labbé, “Antigenic strain diversity predicts different biogeographic patterns of maintenance and decline of antimalarial drug resistance,” eLife, 12:RP90888, 2024.