For early-career researchers, exchanging ideas with top scientists in their research field is crucial. It allows them to benefit from senior researchers’ experience, understand current research trends and deepen their own knowledge through dialogue. The ELLIS PhD & Postdoc Program connects PhD students and postdocs from all over Europe with top experts in all areas of machine learning – both in research and industry. The program also enables its participants to spend six months at a partner institution abroad. PhD student Tobias Lorenz is currently enjoying the benefits of the ELLIS PhD & Postdoc Program in Oxford. Tobias usually works at the CISPA Helmholtz Centre for Information Security, in the research group of Professor Dr. Mario Fritz, who also coordinates the ELSA Network of Excellence. ELSA researchers like Fritz support and expand the PhD & Postdoc Program of ELLIS.
Hello Tobias. You are a participant in the ELLIS PhD-Program and you are currently in Oxford. Could you tell me more about the program and what exactly you are up to in Oxford?
The basic idea of the ELLIS PhD & Postdoc Program is to strengthen the connections between research groups across Europe. As a participant, I have two supervising professors from two different European countries who give me input on my research work. Part of the program is also a stay of at least six months at a partner institution – in my case, this is the University of Oxford. Here I continue to work on my doctoral thesis. My supervisor in Oxford is Professor Marta Kwiatkowska. She is an expert in my research area: the certifiable robustness of machine-learning systems.
Can you please explain this in more detail?
Ok, I’ll try (laughs). Neural networks are at the core of many machine learning systems, but they have not been very robust to date. If the input, i.e. the data we feed the system, is changed just a little, it often impacts the output. This means that attackers can easily manipulate the systems with slightly altered data. For example, if they add a small amount of noise to an image, even if the noise they add is invisible to the human eye, the system is often no longer able to understand what is depicted in the image. We want to make the systems more robust so that they can handle such small changes better and still deliver reliable results.
Two different tracks are on offer in the ELLIS PhD & Postdoc Program: the Academic Track and the Industry Track. Participants can choose between spending six months at another research institution or an industry partner. You have chosen the Academic Track. Are you planning to stay in the academic world?
At the moment at least, I feel very comfortable in academia because there is so much freedom in science. Also, I find foundational research very exciting. But I could also picture myself working in a research department in industry later on. It remains to be seen what exactly will happen after my PhD.
How did you learn about the ELLIS PhD & Postdoc Program and how does the application process work?
The supervisor of my Bachelor’s thesis drew my attention to the program at some point. When I did my Master’s degree at ETH Zurich, my professor there also encouraged me to apply. There is a central application process via ELLIS, in the course of which applicants are matched with professors throughout Europe who are suitable for collaboration on the respective PhD topic. Both sides can then choose who they want to work with. That’s how I came to join Mario’s group at CISPA. I also work together with Mario in the ELSA Network of Excellence, and I found my second supervisor, Marta, through the ELLIS program.
Was working in the ELSA-Network a requirement for participation in the ELLIS PhD & Postdoc Program?
No, ELSA did not even exist then. Working in one of the networks of excellence might help open doors to the ELLIS program, but it is not a requirement. For me working in ELSA is very interesting because it involves a lot of research on topics that are very close to my subject. ELLIS’s scope is much broader and deals with machine learning topics across the board. ELSA, on the other hand, focuses on the security and robustness of systems. So, I benefit from both networks.
How exactly did ELSA help you to realize your stay in Oxford?
ELSA has a mobility program that is designed precisely for this purpose. I can get a part of my travel expenses reimbursed through this program. My PhD position is also funded by ELSA. ELLIS itself does not have any positions, but it cooperates with many universities and research institutions and builds on their programs.
What did you take away from the ELLIS PhD & Postdoc Program? Would you encourage other young researchers to apply?
In addition to providing the opportunity of working here in Oxford, ELLIS also organizes events such as the ELLIS Summer School and the Doctoral Symposium, which I have already attended twice. That was great because I met the other doctoral students. At these events, a lot of exchange takes place and some new ideas emerge. The events are organized by the various ELLIS units, by doctoral students for doctoral students, and they offer a mix of lectures by professors and PhD candidates on a predefined, overarching topic. And there is, of course, lots of exchange between the doctoral students. So yes, even if there is competition for places: You meet top researchers and build up a network. If this is something you are interested in, you should apply.
Have you already planned specific collaborations with researchers you met through ELSA or ELLIS?
There are many ideas already. But I’ll have to see how they can be put into practice.
Tobias, thank you very much for the interview.
About ELLIS:
ELLIS – the European Laboratory for Learning and Intelligent Systems – is a pan-European AI network of excellence which focuses on fundamental science, technical innovation and societal impact. Founded in 2018, ELLIS builds upon machine learning as the driver for modern AI and aims to secure Europe’s sovereignty in this competitive field by creating a multi-centric AI research laboratory. ELLIS wants to ensure that the highest level of AI research is performed in the open societies of Europe and follows a three-pillar strategy to achieve that.