New Perspective in Nature Methods: Advancing Primate Behavior Analysis in the Wild

A recently published Perspective article in Nature Methods, the result of a highly collaborative effort by many members of our SFB, offers a comprehensive overview of computer vision methods for analyzing primate behavior in natural environments. While existing tools have transformed behavioral analysis in laboratory settings, this article addresses the critical gap: how to robustly detect, track, and identify multiple animals, and how to interpret their behavior under the complex, dynamic conditions of the field.This interdisciplinary approach aims to contribute to the ongoing development of methods that are both practical and scalable, with the potential to enhance primate behavior research in natural settings. Our team assesses the limitations of current approaches and proposes a unified, video-centric framework that integrates all key tasks, moving beyond fragmented solutions. By emphasizing data-centric and effort-efficient learning strategies, the article provides a roadmap for scaling behavioral research without exhaustive manual annotation. The authors suggest shifting the focus from tool-specific challenges to holistic, scalable methodologies for real-world behavior analysis.

To the Publication in Nature Methods

Final Panel Discussion: Empowering First Generation Academics

On March 12, we hosted our final panel discussion on how to increase diversity in Academia. The topic was “First Generation Academics – How the Family Background Influences the Career Paths of Young Female Scientists." Our expert panelists – Zurna Ahmed, Holmer Steinfath, Britta Korkowsky, Charlotte Prauß and Ann-Kristin Kolwes – openly discussed the challenges faced by young researchers from non-academic backgrounds. These challenges include not only financial barriers but also self-doubt, uncertainty in the academic environment, and lack of understanding from their own families. But the panel also talked about seemingly simple yet significant hurdles in academic life, such as: How do I find a supervisor? Can I approach professors directly? How do I become a student assistant?

New Publication in PNAS: Understanding Local Learning in Neural Networks

How do individual elements in a neural network contribute to solving complex tasks? While both biological and artificial neural networks achieve remarkable performance, their local learning dynamics remain poorly understood. A new study by Michael Wibral, Viola Priesemann, and colleagues, published in PNAS, introduces a novel framework to describe local learning goals using principles from information theory. The researchers present infomorphic networks, which define learning objectives at the level of individual neurons through a parametric approach based on Partial Information Decomposition (PID). This allows them to unify different learning rules and tasks—including supervised, unsupervised, and memory learning—within a single theoretical framework. By making local learning dynamics more interpretable, infomorphic networks help bridge the gap between theoretical neuroscience and artificial intelligence.

To the press release

Link to the paper