Statistiskt seminarium

Abstrakt finns enbart på engelska: Machine learning is a powerful new tool for making sense of high-dimensional signals (images, waveforms, etc.) that humans struggle to process. In medicine, this is already starting to uncover useful patterns in data - but turning these facts (e.g., this patient is at high risk) into discoveries (e.g., what is the nature of a high-risk patient’s physiology and how can we fix it) has so far proven elusive. I’ll present two examples of new empirical findings, grounded in large medical imaging datasets, and some preliminary insights on how they can drive theoretical understanding. This ‘hybrid’ discovery process - where new facts produced by ML prompt humans to generate and test hypotheses - will be fundamental to an emerging science of medicine, grounded in data and computation.