This understanding is often formalized through the construction of (hopefully general) models, thereby capturing the information contained in different data sources. If successful, one can deploy these constructions to tackle inverse problems of different kinds, prediction, clustering and other machine learning tasks, and much more. 

My work is driven by the design of robust representation learning and machine learning systems, both from a theoretical standpoint as well as with precise applications in biomedical sciences. This is materialized in projects around multilayer sparse modeling and its theoretical underpinnings, the design of more robust deep learning models and their application to cancer detection in digital pathology, and more.