Representation learning of animal behavior Predicting Zebrafish Behavior and Internal States using Computational Models

November 8, 2024

Animal behavior shows significant variability, with genetically identical animals exposed to the same stimuli often responding differently. Often, variability is associated with variables that cannot be directly observed. These variables can be either external (e.g., temperature changes, group dynamics) or internal (e.g., fatigue, hunger, alertness), each introducing variability—or "noise"—in distinct ways. These "hidden" variables are often referred to as internal states, and understanding and measuring them is of interest for behavioral neuroscience to better understand the hidden mechanisms driving observable behaviors. The zebrafish (Danio rerio) has emerged as a valuable model organism for studying these internal states, due to its genetic similarity to humans, transparent larval stage, and well-defined swimming behavior. High-throughput, automated techniques that now allow for precise tracking of animal movements, have facilitated the measurement of behavioral patterns in zebrafish, allowing for the development of large datasets of behavior. This research proposes the creation of a latent variable model based on zebrafish behavioral data to predict future actions based on past sensory inputs and actions. Additionally, the model will be extended to assess the behavioral impact of pharmacological interventions. This work could potentially aid drug screening and contribute to neuroscience and preclinical research.

Authors in the community:

• 

  G

  Gonçalo Goulart Oliveira

  ist195582

Supervisors of this institution:

• 

  A

  Arlindo Manuel Limede de Oliveira

  ist12282

electrical-engineering-electronic-engineering-information-engineering - Electrical engineering, electronic engineering, information engineering

eng - English

Instituto Superior Técnico