Artificial Intelligence & Machine Learning technology are playing a key role in better understanding pandemics. More specifically, AI can support decision makers in taking the right actions to handle a pandemic. We discussed with Dr. Pieter Libin, Professor at the AI lab of the Vrije Universiteit Brussel, how Artificial Intelligence, and more precisely, Reinforcement Learning, a subfield in Machine Learning, can help to take optimal decisions to better predict & better guide the mitigation of epidemics.
What is the role of AI & Machine Learning in fighting pandemics?
In our Business & Decision Expert Podcast, we strive to answer questions in the domain of Data & Digital. For this episode, we will deep dive into Artificial Intelligence & Machine Learning and their used in guiding the mitigation of epidemics.
What we will discuss in this podcast episode:
– What kind of data is captured and how are they collected?
– What are the predictions made by Artificial Intelligence with regard to pandemics?
– What are the Machine Learning techniques used to fit with this data?
– Would it be possible to use Machine Learning to prevent the next pandemics?
Gerrit Denayer, Business & Decision Marketing Director, received Pieter Libin, Professor at the AI lab of the Vrije Universiteit Brussel, to cover this topic:
Listen to the full podcast:
Gerrit: What did you try to achieve with AI in regard to pandemics? Is it the prediction when it would come? Is it the prediction how it would spread? Or is it the anticipation how we could solve it or any combination of that?
Pieter: There are two parts about using AI or Machine Learning and because these two terms can be used interchangeably. First of all, we need to fit models to data. So, we make epidemiological models, and this can be very fine grained, where we explicitly model each individual of the population of Belgium, or they can be more coarse-grained, where we actually compartmentalize the population.
So, these models have a structure, but we still need to fit them towards data. Hospitalization data is typically used because this is a very stable data source. Using this data, machine Learning techniques or statistical frameworks can be used to fit the models to such data but also to quantify the uncertainty that is present given this data points. Especially at the start of the epidemic, there was still a lot of uncertainty in these parameters and taking this into account is also very important to make future predictions.
Reinforcement learning is all about optimizing policies. A policy is, for example, something that the government can implement to try to mitigate the effects of an epidemic or pandemic.Pieter libin
Gerrit: How would the purely statistical model be different from what you said in the beginning, the reinforcement learning?
Pieter: So, Reinforcement learning is all about optimizing policies. So, a policy is, for example, something that the government can implement to try to mitigate the effects of an epidemic or pandemic. In the context of COVID, that could be: how to execute a lockdown? Do we lock the entire society? Do we only close schools? Do we avoid that people go to work? Or do we do we do a combination of all these different aspects. These decisions can be formulated as a public health policy. And reinforcement learning techniques are all about trying to learn an optimal policy, but such reinforcement learning frameworks need simulation model to exercise because a reinforcement learning agent will learn by trial and error. And of course, this is something that you cannot do in the real world. So that is something that we do in a simulator, or an epidemiological model.
Business & Decision Expert is a podcast where we talk about Data, AI & Digital topics. We receive every month Data & Digital experts to discuss how Data is revolutionizing our daily lives and we strive to answer business question related to digital transformation. We aim to gather a community of Data passionate people around our podcast where everyone can learn something in a relaxed and fun way.
You can listen to us on Soundcloud, Apple Podcast, Spotify or other podcast platforms.