Until recently, measuring animal behavior implied either relying on observational studies, or on overly simplified settings. Leveraging advances in machine learning, it lately became common practice to track multiple body parts over time, without physical markers. This thesis aims to (1) develop novel deep clustering algorithms to explore the behavioral repertoire of animals using motion tracking, (2) deploy them in an open-source package, and (3) use them to characterize a real world animal model.
«
Until recently, measuring animal behavior implied either relying on observational studies, or on overly simplified settings. Leveraging advances in machine learning, it lately became common practice to track multiple body parts over time, without physical markers. This thesis aims to (1) develop novel deep clustering algorithms to explore the behavioral repertoire of animals using motion tracking, (2) deploy them in an open-source package, and (3) use them to characterize a real world animal mod...
»