Viruses that cause latent infectious diseases (e.g. HIV, hepatitis C, papilloma) pose an important threat to public health. The most efficient way to mitigate epidemics of latent infectious diseases is by prevention. However, the complex dynamics of such epidemics render the development of effective and efficient prevention strategies challenging. To support decision making with respect to prevention strategies, epidemiological models are frequently used.

There are two main types of epidemiological models that are frequently applied: compartment models, which divide the population into discrete homogeneous states (i.e., compartments) and describe the transition rates from one state to another, and individual-based models that explicitly represent all individuals and their connections, and simulate the spread of a pathogen among these individuals. Since preventive mechanisms are specifically targeted towards individuals, individual-based models allow for a more accurate evaluation of preventive strategies.

In this research project, we will extend the state-of-the-art of individual-based models to allow researchers to study large-scale (i.e. a country) epidemics of latent infectious diseases. An important aspect is to correctly model the structure and dynamics of the social networks over which the pathogen is able to spread. To evaluate our methods, we will construct a model of the HIV epidemic in the Lisbon metropolitan area. We will validate the outcomes of this model based on a dataset that provides a dense sampling of the Portuguese epidemic.

To learn preventive strategies in an individual-based model, we will investigate the use of reinforcement learning. We plan to represent preventive strategies as a set of agents that are able to execute preventive measures in an individual-based model. Through reinforcement learning, we will optimize the preventive agents’ behavior and thereby learn an optimal preventive strategy.

Research topics:

• Computational Biology
• Learning in Multi-Agent Systems
• Reinforcement Learning