Tuesday, 25 February 2025 18:43

Basic research: deep learning network analyzes thinking processes Featured

A research team from Würzburg, Vienna, London and Switzerland have developed a geometric deep learning network called MARBLE, which helps to analyze thinking strategies of the human brain. MARBLE can recognize a common structure in the thinking of two individuals or species without ignoring the unique “language” of each brain.


Dr. Robert Peach, physicist and computational neuroscientist at the Department of Neurology at the University Hospital of Würzburg (UKW), and Prof. Adam Gosztolai, mathematician at the Medical University of Vienna, have developed the computer program MARBLE together with researchers at Imperial College in London and the École Polytechnique Fédérale de Lausanne (EPFL) in Switzerland.

The researchers tested MARBLE on artificial neural networks, simulated systems and real brain data from primates and rodents. They found recurring patterns associated with thought processes such as decision-making or adaptation to new situations.  Instead of examining all neurons individually, MARBLE looks at sections of brain activity and compares them between different species and tasks. The program analyses temporal changes in the signals and can thus detect subtle differences between thinking strategies.

Original publication:
Gosztolai, A., Peach, R.L., Arnaudon, A. et al. MARBLE: interpretable representations of neural population dynamics using geometric deep learning. Nat Methods (2025). https://doi.org/10.1038/s41592-024-02582-2

For more information:
https://www.uni-wuerzburg.de/aktuelles/einblick/single/news/auf-mustersuche-im-gehirn/