Fueled by enhanced computational power, data availability, and commercial reward, machine learning systems now approach, and sometimes surpass, human capabilities to recognize images, play games, and predict/simulate behavior. This unlocks numerous applications—some, like self-driving vehicles or medical decision-making, with direct impact on human lives. Despite extensive empirical progress, our theoretical account of modern machine learning is premature and remains a highly active area of research. The FIND group works to develop the theoretical foundations of statistical learning theory and builds on them to progress the development of learning algorithms that are accurate, efficient, robust, private, and fair. In the long term, these will unlock invaluable societal benefits, from better healthcare to safer roads and improved crisis management.
Specific research areas: Computational medical imaging, computational neuroscience, generative modeling, machine learning in medicine, optimal transport theory, statistical inference, statistical learning theory.