Department of Information Technology

Research profile: Thomas Schön

Thomas Schön

Thomas Schöns research is about Machine Learning, about building mathematical models based on data and giving the computer an ability to learn things that it is not specifically programmed for. Photo: Kajsa Örjavik

"I am somewhat of an entrepreneur"

Thomas Schön has built a successful research team within Machine Learning at the Department of Information Technology. There it is boiling with activities in many different research areas, much thanks to Thomas's ability to put together collaborations and push through projects.

- My research is about Machine Learning, which is part of AI, artificial intelligence, says Thomas Schön, professor of Automatic Control at the Department of Information Technology at Uppsala University. It is about building mathematical models based on data and giving the computer an ability to learn things that it is not specifically programmed for.

Thomas' interest in Machine Learning was raised a few years after his dissertation. There were two main reasons for this.
- First, the mathematics that I master and think is fun are used, and second, the mathematics are used to solve a much broader set of problems than I was used to, so I got access to a variety of new and exciting applications. I also thought that Machine Learning would have a large impact in society, but the most important reason was probably that I simply couldn't help but go into the area.

There are many use cases for Machine Learning. And a lot happens in Thomas's research team. In fact, it seems to be boiling with activities.
- Something that we are working with right now is automated interpretation of ECG. It is a project where we collaborate with researchers in Brazil. In fact, our auto-interpretation is actually better than real doctors in finding five of the six most common heart defects nowadays. We have just published fresh results from that research!

Thomas says he has an endeavor to deliver the best basic research and the most relevant applied research. The applied research is done together with companies or applied research groups.
- One example of this is the research on the automated interpretation of ECG where cardiologists in Brazil have collected data for ten years and has created a research group, explains Thomas. A guest doctoral student at the department comes from that group. He presented the project and wondered if I wanted to help. We helped with the mathematical models (deep learning in this case) and they contribute with the clinical knowledge and so we get a good use of our knowledge in a common application.

Thomas's research team develops mathematical models and examines their properties and how they can be used. Which application they work on and continue to research within is usually determined in a fairly random manner. As long as the research can benefit society.
- A collaboration can start because I have a meeting with someone to talk about something completely different. In fact, I expose myself to situations and meetings to see what ideas can arise.

One of the focus areas for Thomas's research is medicine.
- It started with me getting good contacts with doctors through a leadership course that I attended. Uppsala is good in this way by being close to the University Hospital and doctors who are doing research.
- Then, when my mother got cancer, it affected me personally, so I wanted to find some way to help making cancer-care better.
When the company Elekta contacted Thomas and wanted help with the development of a new machine for radiation therapy, it felt right.

Another new research area within Machine Learning that Thomas colleague Dave Zachariah has begun to work on is about causality - to systematically find cause-effect relationships from human knowledge and large data sets.
- It feels very exciting methodically and application-wise, says Thomas. Finding causal relationships, such as between behavior and illness.

Yet another result from Thomas's research team is a new programming language. The language is specially built to give more people access to using the rather complicated algorithms Sequential Monte Carlo (also known as particle filter). Using these algorithms you can solve many kinds of problems.
- It can be used in virtually all scientific areas, says Thomas. We have recently tried to use it in phylogenetics, which is about mapping how species develop and become extinct. On this application we work with Fredrik Ronquist who is a Professor at the Swedish Museum of Natural History. It has also been used for epidemiological studies, such as the zika virus.

The main theme in Thomas's research is dynamic systems, things that change over time and how to mathematically describe it using models that reason with uncertainty.
The research can help solving many different kinds of problems and has great relevance for most research areas where large amounts of data are handled. There are collaborations with medicine, mathematics, physics, as well as peace- and conflict science, to name a few.
- The biggest challenges are finding staff and time. I am constantly looking for competent researchers to recruit.

Thomas Schön's biggest achievement, according to himself, is building a strong team within Machine Learning.
- Five years ago I was alone and today we are over 20, Thomas says. From the beginning I had a strong desire to create a team to work in. I do not want to sit and research by myself, I get bored.

Thomas believes that one of his success factors is that he is somewhat of an entrepreneur in the University environment.
- I am good at putting together flexible collaborations, picking up people from all over the world who are most suitable for doing a certain thing and making sure that the collaborative work actually gets done.
The next goal is to try to get more talented employees at senior level to the team and to get the knowledge out so that it can be used by others.

Thomas also teaches and it is something that he likes and thinks is important.
- It's really exciting to create an interest among the students. In addition, it is refreshing to work with young people who are not locked into a certain way of thinking, but instead think freely, question and come up with new ideas. It's a luxury in this job!

Facts - THOMAS SCHÖN

Age: 41
Title: Professor of Automatic Control
Education: Master of Science in Applied Physics and EE, Licentiate of Engineering in Automatic Control and Bachelor of Science in Business Administration.
Place of residence: Uppsala
Family: A lovely family consisting of parents, a brother and close friends around the world.
Leisure time activities: Spends a lot of time in the nature in different ways, training a lot, skiing (especially ski touring) hiking and hunting.
Currently: I’m trying to initiate solid research and education in artificial intelligence / machine learning in Sweden via the research project WASP (Sweden's largest individual research project so far).
Listening to: Wise people (via real-time meetings, books, podcasts, etc.). Has a broad taste of music, but is worthless in remembering artists.
Hidden talent: Can climb vertical walls of ice.
Strength: Creates opportunities.
Weakness: Time optimist.
Dream project: To build a log cabin in the mountains where there is a lot of snow and exciting skiing and hiking to be done. I like the area around Abisko and the parts of northern Norway I have visited so far.

Updated  2019-01-14 15:47:45 by Kajsa Örjavik.