Senior Group Leader and Head of Computation and Theory
The aim in my research is to develop new machine vision and learning tools, methods, and frameworks that have a big impact on biology and how biology is done. Working with my colleagues at Janelia, I try to discover problems for which machine vision and learning-based solutions would benefit many labs. I engineer solutions that jointly consider the entire context of the problem. For example, how should the data be collected -- can the computer vision problem be simplified by e.g. — carefully controlling lighting and camera angle? How do we efficiently obtain the careful human annotations required for supervised learning? What is the state of technology -- what problems and problem formulations are robustly solvable using modern machine vision and learning techniques? Which algorithms work in which situations, and where do we need to do algorithm development? How do we make such software intuitive for biologists without deep knowledge of the underlying algorithms? Should we modify our mathematical criterion for success — which kinds of errors matter more than others? Are there problematic biases introduced by errors? Does the system work well enough to do meaningful science? What should be solved by biologists versus by these automated systems? And, most challengingly, how do we gain biological insight and understanding from the big data sets that these machine learning approaches enable? As machine vision and learning research has accelerated in the past decade, it is the last problem that I believe is the next frontier for computational biology research. Our lab is exploring new methods for combining modeling with machine learning as well as methods for data visualization and exploration.