Analysis of complex and subtle behavior enabled by unsupervised deep learning

• Applicant: France Rose, Ph.D.
• Subject Area: Cognitive, Systems and Behavioural Neurobiology ; Medical Informatics and Medical Bioinformatics
• Term: since 2025
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 556006703

Studying freely moving animals is essential to understand how animals behave and make decisions – e.g. when they escape predators, find mates, or raise their young – in an undisturbed manner. Although animal behavior has been studied for decades, it has only recently become possible to measure movement mechanics in a high-throughput manner due to advancements in cameras, motion capture systems, and pose estimation methods. In parallel, powerful Deep Learning (DL) methods emerged for handling sequential data, including text (natural language processing) and human movement (action prediction). However, DL is only since recently used to study animal behavior, and primarily as detection methods like pose estimation. Here, I propose to develop new, flexible DL methods to quantitatively describe animal movement and its link to neuronal activity. I will explore unsupervised, self-supervised, and transfer learning to study behavior without relying on heavy manual labeling. I will also leverage biological perturbations (genetic, pharmacological, social interactions) and create a general framework to improve our understanding of the behavior diversity in control and perturbed conditions. Such a fine description of unconstrained and unconditioned behavior could further be applied to other situations, including patients suffering from musculoskeletal and neuronal disorders, bearing a great potential for improved diagnostics and new therapeutic strategies.

• DFG Programme: Emmy Noether Independent Junior Research Groups