On the fourth day of the Track AI MIDSEA Summer School 2025 series of activities, participants explored the application of Reinforcement Learning (RL) in the context of infectious disease intervention. The first session began with an introduction to the basic concepts of RL, where participants learned about the main structure of RL agents, including the state, action, reward, and policy components. This explanation was presented visually using interactive simulations to facilitate understanding.
The material continued with an in-depth presentation titled “Reinforcement Learning for Infectious Disease Interventions.” In this session, participants are introduced to modeling intervention systems as Markov Decision Processes (MDPs), where RL agents formulate intervention policies through repeated interactions with a simulation environment that represents the dynamics of disease spread. The goal is to minimize the number of infections while considering the socio-economic impact of the policy.
The speaker presented the implementation of Deep Q-Learning and Policy Gradient in epidemiological scenarios, and discussed key challenges such as sample inefficiency and uncertainty in epidemiological models. The discussion also highlighted the importance of integrating real-world data and interpretability in the application of RL for public policy making.
In addition to gaining theoretical understanding, participants also had the opportunity to directly simulate RL algorithms using Python notebooks, enabling them to explore the dynamics of agent learning in the context of epidemic control.
The fourth day of activities concluded with an open discussion and question-and-answer session, where participants and presenters engaged in dialogue regarding the potential for further research collaboration and broader applications of reinforcement learning in the public health sector.
Keywords: MIDSEA, Modelling, Infectious Diseases Modelling
Author: Chyntia Meininda Anjanni
Photo: Lucetta Amarakamini