Showing 1–2 of 2 results for author: Čutura, G

Deep Demixing: Reconstructing the Evolution of Network Epidemics

Authors: Boning Li, Gojko Čutura, Ananthram Swami, Santiago Segarra

Abstract: We propose the deep demixing (DDmix) model, a graph autoencoder that can reconstruct epidemics evolving over networks from partial or aggregated temporal information. Assuming knowledge of the network topology but not of the epidemic model, our goal is to estimate the complete propagation path of a disease spread. A data-driven approach is leveraged to overcome the lack of model awareness. To solv… ▽ More We propose the deep demixing (DDmix) model, a graph autoencoder that can reconstruct epidemics evolving over networks from partial or aggregated temporal information. Assuming knowledge of the network topology but not of the epidemic model, our goal is to estimate the complete propagation path of a disease spread. A data-driven approach is leveraged to overcome the lack of model awareness. To solve this inverse problem, DDmix is proposed as a graph conditional variational autoencoder that is trained from past epidemic spreads. DDmix seeks to capture key aspects of the underlying (unknown) spreading dynamics in its latent space. Using epidemic spreads simulated in synthetic and real-world networks, we demonstrate the accuracy of DDmix by comparing it with multiple (non-graph-aware) learning algorithms. The generalizability of DDmix is highlighted across different types of networks. Finally, we showcase that a simple post-processing extension of our proposed method can help identify super-spreaders in the reconstructed propagation path. △ Less

Submitted 10 June, 2023; originally announced June 2023.

Comments: arXiv admin note: substantial text overlap with arXiv:2011.09583

Deep Demixing: Reconstructing the Evolution of Epidemics Using Graph Neural Networks

Authors: Gojko Cutura, Boning Li, Ananthram Swami, Santiago Segarra

Abstract: We study the temporal reconstruction of epidemics evolving over networks. Given partial or aggregated temporal information of the epidemic, our goal is to estimate the complete evolution of the spread leveraging the topology of the network but being agnostic to the precise epidemic model. We overcome this lack of model awareness through a data-driven solution to the inverse problem at hand. In par… ▽ More We study the temporal reconstruction of epidemics evolving over networks. Given partial or aggregated temporal information of the epidemic, our goal is to estimate the complete evolution of the spread leveraging the topology of the network but being agnostic to the precise epidemic model. We overcome this lack of model awareness through a data-driven solution to the inverse problem at hand. In particular, we propose DDmix, a graph conditional variational autoencoder that can be trained from past epidemic spreads and whose latent space seeks to capture key aspects of the underlying (unknown) spreading dynamics. We illustrate the accuracy and generalizability of DDmix and compare it with non-graph-aware learning algorithms through numerical experiments on epidemic spreads simulated on synthetic and real-world networks. △ Less

Submitted 18 November, 2020; originally announced November 2020.