Professor portrait: Veronika Cheplygina improves the field of machine learning through meta-research
On 10 June 2025 at 14:30, Professor Veronika Cheplygina will present her inaugural lecture in Auditorium 0 at the IT University of Copenhagen. The lecture is entitled: “Not real research”.
Veronika CheplyginaAbout ITUEventsProfessor portrait
Written 26 May, 2025 07:37 by Mette Strange Mortensen
Exactly 10 years after she defended her PhD, Veronika Cheplygina will give her inaugural lecture as professor at the IT University of Copenhagen. She started her academic career in machine learning, but already during her PhD she started questioning the way research was done and wondered if research was just optimised for publication.
“My PhD research was about specific types of machine learning, which could be used for several applications, including medical imaging, which I focused on after my PhD. While reading papers about it, I saw various patterns about how people were doing research. For example, I saw that people were introducing essentially the same method but calling it something entirely different, perhaps for the sake of novelty”, says Veronika Cheplygina, “so you would never find the literature on both if you didn't try to search for synonyms, or more broadly, look at the assumptions used to design a method. What I was noticing was these problems with how the research might be optimised for publication rather than for generalisable results.”
She started looking further into this and similar phenomena and wrote several survey papers about this. In recent years, her research has shifted towards research on what people are investigating and why. For example, she is interested in what kind of data people use and if the quality of the data and the evaluation practices actually justify the conclusions of the research.
“What I find interesting about my work is that not a lot of people want to work with data. I think this is in part because of the emphasis placed on novelty and state-of-the-art results when publishing at conferences. Essentially, you can download some publicly available data, and without understanding the data, immediately get started with optimizing your method until you get good results. I guess it is more difficult to look across multiple different studies and see if the conclusions are the same,” says Veronika Cheplygina, “and this is also a more difficult way to publication, than introducing a novel method. I find it interesting that people do agree that such studies are important, but that they often do not do it”.
Veronika Cheplygina’s inaugural lecture will be held on 10 June 2025 at 14:30 in Auditorium 0 at the IT University of Copenhagen. The lecture is entitled: “Not real research”. This is a reference to the reception of the work that Veronika Cheplygina does.
“In a previous job I was asked when I was going to start doing real research. The implication was that writing surveys instead of proposing novel methods was not real research. It really left an impression,” says Veronika Cheplygina.
Her “not real research” has now made her a professor at ITU. However, she is not afraid to share when she is failing. Most researchers share when they get a new grant or publish in big journals. So does Veronika Cheplygina, but she also talks about all the papers, promotions and grants she did not get:
“I mean researchers are just people. So, it doesn't make sense to me to pretend to be like this person on the pedestal,” says Veronika Cheplygina, “as a postdoc I saw that a Professor at Princeton University shared a CV of failures, and thought at that point, that mine was already longer. When I shared it, I got so many positive responses. It somehow felt more impactful than publishing a paper that maybe ten people would read. I continued updating it and have also interviewed other researchers on their failures.”
Looking into diversity in teams and algorithms
Even though Veronika Cheplygina’s research has been called not “real”, and she has published her failures, she has also managed to succeed. She came to ITU in 2021 during the COVID lockdown and for the first six months had to do her research from home. This gave her time to read and write a lot of research about research. This research recently landed her a grant from the Novo Nordisk Foundation.
“I am doing a project called CHEeTAh: CHallenges of Evaluating Teams and Algorithms, where my team and I are investigating how the practice of machine learning competitions, often hosted at conferences or in education, could potentially exclude researchers and students who are not eager to compete or simply do not have the resources to do so. This project aims to explore inclusive ways of organizing machine learning competitions by investigating whether competitions that require algorithms to succeed on multiple evaluation metrics can lead to better algorithms and more diverse teams,” says Veronika Cheplygina, “I hope that this project will also leads to tools that can be used at conferences or in classrooms. I am not only finding problems, I am also proposing solutions.”
You can read an article about the CHEeTah: CHallenges of Evaluating Teams and Algorithms project here.
Theis Duelund Jensen, Press Officer, phone +45 2555 0447, email