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

Epidemiologists model the dynamics of epidemics in order to propose control strategies based on pharmaceutical and non-pharmaceutical interventions (contact limitation, lock down, vaccination, etc). Hand-designing such is not trivial because number possible difficulty predict long-term effects. This task can be cast as an optimization problem where state-of-the-art machine learning algorithms deep reinforcement learning, might bring significant value. However, specificity each domain -- epidemic modelling or solving requires strong collaborations between researchers from different fields expertise. why we introduce EpidemiOptim, a Python toolbox that facilitates epidemiology optimization. EpidemiOptim turns epidemiological models cost functions into problems via standard interface commonly used by practitioners (OpenAI Gym). Reinforcement Q-Learning with neural networks (DQN) evolutionary (NSGA-II) are already implemented. We illustrate use find optimal policies for dynamical on-off lock-down under death toll economic recess using Susceptible-Exposed-Infectious-Removed (SEIR) COVID-19. Using its interactive visualization platform Jupyter notebooks, epidemiologists, others (e.g. economists) easily compare models, costs address important choices made health decision-makers.