EpidemiOptim: a Toolbox for the Optimization of Control Policies in Epidemiological Models - Archive ouverte HAL Accéder directement au contenu
Article Dans Une Revue Journal of Artificial Intelligence Research Année : 2021

EpidemiOptim: a Toolbox for the Optimization of Control Policies in Epidemiological Models

Résumé

Modelling the dynamics of epidemics helps proposing control strategies based on phar-maceutical and non-pharmaceutical interventions (contact limitation, lock down, vaccina-tion, etc). Hand-designing such strategies is not trivial because of the number of pos-sible interventions and the difficulty to predict long-term effects. This task can be castas an optimization problem where state-of-the-art machine learning algorithms such asdeep reinforcement learning, might bring significant value. However, the specificity ofeach domain – epidemic modelling or solving optimization problem – requires strong col-laborations between researchers from different fields of expertise. This is why we intro-duce EpidemiOptim, a Python toolbox that facilitates collaborations between researchersin epidemiology and optimization. EpidemiOptim turns epidemiological models and costfunctions into optimization problems via a standard interface commonly used by optimiza-tion practitioners (OpenAI Gym). Reinforcement learning algorithms based on Q-Learningwith deep neural networks (dqn) and evolutionary algorithms (nsga-ii) are already im-plemented. We illustrate the use of EpidemiOptim to find optimal policies for dynamicalon-off lock-down control under the optimization of death toll and economic recess using aSusceptible-Exposed-Infectious-Removed (seir) model for COVID-19. Using EpidemiOp-tim and its interactive visualization platform in Jupyter notebooks, epidemiologists, op-timization practitioners and others (e.g. economists) can easily compare epidemiologicalmodels, costs functions and optimization algorithms to address important choices to bemade by health decision-makers. Trained models can be explored by experts and non-experts via a web interface.
Fichier principal
Vignette du fichier
12588-Article (PDF)-27614-1-10-20210723.pdf (4.95 Mo) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-03099898 , version 1 (06-01-2021)
hal-03099898 , version 2 (29-11-2021)

Identifiants

Citer

Cédric Colas, Boris P. Hejblum, Sébastien Rouillon, Rodolphe Thiébaut, Pierre-Yves Oudeyer, et al.. EpidemiOptim: a Toolbox for the Optimization of Control Policies in Epidemiological Models. Journal of Artificial Intelligence Research, 2021, 71, pp.479-515. ⟨10.1613/jair.1.12588⟩. ⟨hal-03099898v2⟩
210 Consultations
232 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More