idw – Informationsdienst Wissenschaft
Nachrichten, Termine, Experten
Nachrichten, Termine, Experten
idw - Informationsdienst
Wissenschaft
What do mosquito populations and physical measurement data have in common? Both lead to a central problem in machine learning: the reliable estimation of class prevalence in the face of changing data. As part of a research stay in Brazil, a scientist from the Lamarr Institute and TU Dortmund University is collaborating with international partners on methods that account for such shifts in distribution and enable more robust AI systems
Foz do Iguaçu, Brazil, April 14, 2026. What connects the spread of mosquito populations with the analysis of physical measurement data? More than meets the eye: In both cases, researchers face the same challenge of making reliable statements about frequencies even as the underlying data changes. This is precisely what Dr. Mirko Bunse, a research associate at the Lamarr Institute for Machine Learning and Artificial Intelligence and AI researcher at TU Dortmund, is working on during a research stay in Brazil. Together with Dr. André Maletzke from the Universidade Estadual do Oeste do Paraná (Unioeste), he is developing methods that systematically account for such shifts.
A common problem across disciplines
The focus is on estimating what is known as class prevalence, that is, the frequency of classes within a dataset. Unlike traditional classification tasks, the goal here is not to correctly classify individual objects, but to precisely determine their relative frequencies. This perspective is crucial in various contexts: In Brazil, smart mosquito traps are used to quantify populations of certain species, providing a basis for public health measures. In physics, the same question arises when analyzing experimental data, such as when determining particle types or in the case of data from neutrino telescopes, which also deal with the relative abundances of different classes.
The problem lies in the shift in data distributions: models are trained under specific conditions and encounter altered data when applied. Without proper adaptation, this leads to systematic biases. The joint research addresses precisely this issue and develops methods that model these shifts and integrate them into the estimation.
An Interdisciplinary Comparison of Methods as a Research Approach
The collaboration brings together different scientific perspectives to address a common methodological challenge. While the Brazilian research group develops its approaches for ecological and health-related applications, Bunse contributes expertise in physical analysis. This gives rise to shared methodological approaches that can be further developed across disciplinary boundaries. “We work with very different types of data, ranging from mosquito populations to physical measurement series. But the underlying question is the same: How can frequencies be reliably determined even when the data changes? This exchange helps us develop methods that are more robust against precisely such changes,” says Bunse. For the Lamarr Institute for Machine Learning and Artificial Intelligence, this form of collaboration is exemplary. It demonstrates that key challenges in machine learning are not tied to individual application areas. This leads to methods that remain stable and transferable even under changing conditions.
Caroline Winter
Lamarr Institute for Machine Learning and Artificial Intelligence
c/o Rheinische Friedrich-Wilhelms-Universität Bonn
Friedrich-Hirzebruch-Allee 6
53115 Bonn
Criteria of this press release:
Journalists, Scientists and scholars
Biology, Information technology, Medicine, Physics / astronomy
transregional, national
Cooperation agreements, Miscellaneous scientific news/publications
English
You can combine search terms with and, or and/or not, e.g. Philo not logy.
You can use brackets to separate combinations from each other, e.g. (Philo not logy) or (Psycho and logy).
Coherent groups of words will be located as complete phrases if you put them into quotation marks, e.g. “Federal Republic of Germany”.
You can also use the advanced search without entering search terms. It will then follow the criteria you have selected (e.g. country or subject area).
If you have not selected any criteria in a given category, the entire category will be searched (e.g. all subject areas or all countries).
