LIP External Colloquium: Johannes Niediek - Reinforcement learning to model complex rat behavior
- Datum: 10.09.2025
- Uhrzeit: 13:00
- Vortragende(r): Johannes Niediek, TU Berlin
- Ort: Max-Planck-Institut für Bildungsforschung, Lentzeallee 94, 14195 Berlin
- Raum: Kleiner Sitzungssaal
- Gastgeber: LIP
- Kontakt: seklindenberger@mpib-berlin.mpg.de
- Rubrik: Vorträge
Johannes Niediek, TU Berlin -
Reinforcement learning to model complex rat behavior
Behavior is fast: in a lab experiment, successfully completing a trial usually requires multiple actions. Nevertheless, animal behavior is often quantified via summary statistics such as trial-based success rates. To understand how the brain controls behavior, a fine-grained yet compact description of behavior is necessary. Here, we compute a principled, rich set of candidate policies spanning the range from uninformative to optimal. We match observed rat behavior to these policies and select the best fitting ones, quantifying the alignment of behavior with task requirements. With this approach, we can document on-going refinement in the performance of a behavioral task for many months following the saturation of trial success rates. Moreover, many neurons in the insular cortex of rats track task alignment with a resolution of a few minutes.
Meeting number: 2741 490 0913
Password: ZuN9htnpF48
Behavior is fast: in a lab experiment, successfully completing a trial usually requires multiple actions. Nevertheless, animal behavior is often quantified via summary statistics such as trial-based success rates. To understand how the brain controls behavior, a fine-grained yet compact description of behavior is necessary. Here, we compute a principled, rich set of candidate policies spanning the range from uninformative to optimal. We match observed rat behavior to these policies and select the best fitting ones, quantifying the alignment of behavior with task requirements. With this approach, we can document on-going refinement in the performance of a behavioral task for many months following the saturation of trial success rates. Moreover, many neurons in the insular cortex of rats track task alignment with a resolution of a few minutes.
Join on-site or hybrid:
https://mpib-berlin.webex.com/mpib-berlin/j.php?MTID=mb9688bc4f67e13fda15a5dc0fe206ef5Meeting number: 2741 490 0913
Password: ZuN9htnpF48