In this short project period, the research will provide the Public Health Agency of Sweden with effective and sustainable decision support for choosing strategies to fight pandemic viruses, such as the Coronavirus. The long-term vision is that artificial intelligence should become a natural part of the authority's models and analyses.
– The research project will help the Public Health Authority of Sweden to develop models that recommend measures for preventing the spread of infections based on simulations of the spread and its consequences, Aron Larsson says.
The project will investigate what effect a strategy, or a combination of strategies, has on the spread of a virus. Aron Larsson explains that the strategies are developed in collaboration with domain experts to determine how they can be combined, when they are best applied in time and under what conditions they should be applied.
Aron Larsson further explains that reinforcement learning as a machine learning method will be tested initially with the simulation model VirSim.
– The integration of deep learning with reinforcement learning can function as a powerful tool towards virus spread prediction. The same technology is behind self-driving cars and also behind machines beating world champion players in games such as AlphaGo, Panagiotis Papapetrou stresses.
– The implementation of such an AI agent can lead to a more efficient design of intervention strategies, and that the strategies will be more robust and better meet the valuation criteria. The simulated strategies will then be evaluated to determine how effective they are in relation to each other based on, among other things, how many intensive care sites and deaths can be avoided and what direct and indirect costs the different strategies incur, Aron concludes.
The project will last until 2021 and is part of Vinnova's initiative "Start your AI journey! Public organizations".