Alexander Mathis: Measuring behavior and modeling the brain with machine learning

Quantifying behavior is crucial for many applications across the life sciences and engineering. Videography provides easy methods for the observation and recording of animal behavior in diverse settings, yet extracting particular aspects of a behavior for further analysis can be highly time consuming and computationally challenging. 

I will present an efficient method for markerless pose estimation based on transfer learning with deep neural networks that achieves excellent results with minimal training data (DeepLabCut). Subsequently, I will present methods that enable robust zero-shot performance and action segmentation. Finally, I will talk about top-down and bottom-up approaches for modeling the sensorimotor system.

Alexander Mathis is an Assistant Professor at EPFL. His group works at the intersection of computational neuroscience and machine learning. Ultimately, his group is interested in reverse engineering the algorithms of the brain, in order to figure out how the brain works and to build better artificial intelligence systems. While doing so, he strives to develop open-source tools for the analysis of animal behavior. In parallel, he develops models and theories of adaptive sensorimotor control. 

Previously, he was a Marie-Curie Fellow at Harvard University and the University of Tübingen. He completed his doctorate training at Ludwig Maximilians University in Munich (LMU), after studying mathematics at LMU.