In the Spotlight: Kerstin Bunte

• Name: Kerstin Bunte
• Country of origin: Germany
• Function at BI, if applicable including research group:  Professor of Machine Learning for interdisciplinary data analysis and head of the intelligent systems group
• Year started working at BI (or previous institute): 2007 as a PhD with Michael Biehl After finishing her PhD left for postdoc positions and fellowships elsewhere, but returned in July 2016 (Rosalind Franklin Fellowship)
• Webpage: https://www.cs.rug.nl/~kbunte/

Can you remember the first day you walked into the building/institute and what your impression was at that moment?

The first day I walked into this building was in 2007 when I started my PhD. They actually just opened the building. I was using the staircase to get to the 5th floor and had to go under some sort of construction. That was my first day. I thought “It is a nice building. But maybe they should open it when it is actually finished”.

What motivates you to do your (kind of) work?

The projects are very interesting, they are really interdisciplinary. You have similar problems, but totally different ways of thinking about them and how to solve them. There is an incredible amount of knowledge, and in CS you may not have heard about it. You see more and more that there is an exchange of knowledge and that brings new perspectives.

What are you most proud of that you have achieved so far? (Can be something scientific, but for instance also a new cooperation, becoming member of something, receiving a grant, being able to help out, arrange something, being part of a panel etc)

I am proud of the VIDI grant to be looking into data limitations. We are currently preparing 3 manuscripts. For these papers researchers from four different fields work together to analyze the problems: from engineering, machine learning, applied mathematics and medical applications. This gives insight into how much still needs to be done and it is really interesting to see how these researchers look deeply into the problems to be solved.

Is there anything you want to achieve or establish with your work? (personal, scientific or societal motivation)

One of the major aims is to do more with less. I am striving for efficiency: how can I work with limited and imperfect data and still do the most. Machine Learning is very aware of working with data imperfections such as noise, but there is much less knowledge for learning in situations where key aspects have not or cannot be measured. These situations are quite well understood in fields such as Engineering, computational Biology and mathematics. In medicine most measurements are break up products of very complex dynamic processes. The 'real' quantities of substrates of interest are typically never accessible directly.

There are certain players in the reaction, I know they are there, but I do not know how many or what they exactly are. You can’t measure everything and there is also individual variability. There are things I would like to know, but I will never be able to measure it.

In order to learn in those situations and overcome the limitation of the data, we include expert knowledge in the form of dynamic system models. This allows us to inform the machine learning process and constrain the otherwise infinite possibilities of explaining the data.

Can you state the societal relevance in your research in just a few words?

I am also involved in projects with medical applications. For example medications for children, and then you have to deal with ethical matters limitations and that may lead to suboptimal treatment. In many medical treatments, children are being treated as tiny adults and that means they are often over- or under-treated.

It would be better to incorporate knowledge and data about the development dependent on the age. Including that in the form of applied mathematics modeling we should be able to predict which treatment doses are likely to perform better and lead to truly personalized treatment.

When you were a kid, did you already envision yourself doing something like this?

I was interested in the similarities of mathematics, music and arts. The preludes of Bach are mathematical transformations to hear, while the drawings of M.C. Escher are some to see. Back in the days I was interested in mathematics, physics and Biology. Luckily there was a new interdisciplinary study field in Bielefeld “Informatics for the Natural Sciences”. It thus allowed me to follow all my interests simultaneously.

Is there anything that was organized at your previous job/institute that might be a good suggestion for our institute?

At Universität Bielefeld every summer they had a festival of sound: Nacht der Klänge. Researchers could express some sort of musical, artistic, cultural or experimental thing, often combined with light in surreal settings open to the public. That was really fun and interesting.

What personal detail can you share that people might not know about you?

I am an open person, but there might be some things people do not know about my past. Besides being a scuba diver educated for recovery and rescue of missing persons (finding corpses) and my participation in annual Triathlons when I was in my 20s, I had some unusual jobs. I had a couple of side jobs starting in high school and during my studies to be financially independent. I started with being a waitress and telling 17 year olds that I am not selling them the beer they want, but rather a Coke if they cannot proof their age with ID. While I was 16 at that time ...

Later I worked for a Software company that developed programs for automatic car/truck fueling systems and for the supermarket chain EDEKA in Germany. And 2 nights a week during my studies I would wash passenger planes, such as Boeing, for Lufthansa and Air Berlin at Paderborn airport.

Surely nobody knows these things!

What do you like and dislike about The Netherlands, Groningen and the Dutch people/culture?

What I do like is cycling. What I don’t like are the house prizes, they are ridiculous. I like it that Dutch people are usually direct ... I am as well. They often praise their flat hierarchies, which is a standard they do not necessarily consistently adhere to ;-).

Interviewer: Tanja van der Woude-van den Houten