First of all, congratulations on receiving tenure! What does reaching this career milestone mean to you?
For most young researchers, tenure is something like the ultimate goal, something that you work toward when starting your career. Along the way, you keep wondering whether it will work out or not. Having achieved it now is a huge relief. I'm happy that it worked out, and I feel much more relaxed. At the same time, it's like many other goals in life: Once you reach them, you start wondering what comes next. Of course, I have plenty of ideas and, above all, I want to do great research. What tenure gives me is the security to focus on the truly important topics together with my research group.
What is it that makes working at CISPA special in your eyes?
Above all, I appreciate the amazing colleagues at the center. We have a great relationship and support each other. I also value the excellent facilities and the freedom we have to implement our projects. We have strong funding too, which allows us to tackle truly significant societal challenges. And the atmosphere is great, this shared spirit of growing the center and making it a success. It’s great to be part of a community that wants to make a difference.
You work in the field of machine learning. What were the topics that you focused on during your tenure-track phase?
That’s actually covered quite well by the ERC Starting Grant that I received in 2023. Our goal is to make machine learning significantly more efficient and robust. We’re incorporating ideas from statistical physics to model neural networks on a different scale and develop a more natural parameterization.
What has been your biggest success at CISPA so far, or what is the moment that you’re most proud of?
I feel like a few things came together in a fortunate way. The first was definitely the ERC Grant. That made a huge difference for me and it also increased our visibility in the community. From that moment on, I knew I was heading toward tenure. The second thing was receiving various collaboration requests. For example, Apple Research approached me with a project proposal, and at the NEURIPS conference, I had talks with an American startup. And third, I’m really proud that I was able to recruit some truly brilliant people who are passionate about our work. For example, I was able to recruit a PyTorch engineer that I met at a friend’s wedding. PyTorch is the environment where most machine learning is developed. That is exactly what we need to make our algorithms truly efficient and useful for the research community. I feel like we now have a group of people who can really make a difference. Of course, we still have to put everything into practice, but the foundation is in place.
What would you like to focus on in your research over the coming years?
It’s a mixture of three things: first, applying Complex Network Theory to Deep Learning, second, the efficiency and performance of relevant algorithms and models and third, gaining a deep understanding of specific scientific domains, i.e. our main application areas. We’re facing two distinct challenges: One is the issue of scaling when dealing with massive datasets and training very large models. The other is extracting valuable insights from small datasets, which tend to be noisy but offer other forms of knowledge that we can leverage to achieve greater data efficiency. My team and I want to work with minimal data and small models. This goes against the mainstream trend in machine learning, which emphasizes the need for massive datasets and large models. Right now, the dominant approach is "as big as possible.” The problem with that is it’s very expensive, energy-intensive, and we can’t use it on mobile devices. That’s why we need alternatives.
Now that you’re tenured, will anything change about your work?
Yes, in the sense that my group will be growing. This means that we will have to find good ways of working efficiently with one another. For example, I won’t be able to supervise quite as closely as I did with only two or three PhD candidates. The group will have to support one other and we will have to distribute responsibility effectively. I won’t be able to look at every detail before deadlines the way that I used to. We’re working on balancing this out so that we can continue to have a lot of fun with more people.
Thank you for your time, Rebekka!
