Authors:

William Retnaraj

Date Created:

Not specified

Course Title:

Professor:

Not specified

About Paper:

Containment strategies for spreading processes, such as epidemics, require a good estimate of key parameters that govern their dynamics. However, accurately identifying these underlying parameters from data is often a challenging task. In this work, we address the problem of parameter identification in epidemiological spreading processes, which is frequently complicated by numerical ill-conditioning inherent to the model structure and the lack of persistence of excitation necessary for the convergence of adaptive learning schemes. To overcome these challenges, we propose leveraging a relaxed property called initial excitation, combined with a recursive least squares algorithm, to design an online adaptive identifier. This identifier learns the parameters of the susceptible-infected-susceptible (SIS) epidemic model from the available knowledge of its states. We provide a proof that the iterates generated by our proposed algorithm minimize a relevant auxiliary weighted least squares cost function. To validate our approach, we conduct numerical case studies encompassing both aggregate-population SIS models and networked SIS models. Through these studies, we demonstrate the convergence of the error in the estimated epidemic parameters and compare our results with those obtained using conventional identification approaches. Specifically, we showcase the successful recovery of parameters and contact network structure for aggregate-population SIS models and networked SIS models, respectively.

Source:

Purdue University / 2023

Topics:

No topics listed

Co-authors:

William Retnaraj