Efstathios (Stathis) D. Gennatas, MBBS AICSM PhD, Research Scientist at the Laboratory for Artificial Intelligence in Medicine and Biomedical Physics, Stanford University School of Medicine
Expert-Augmented Machine Learning (EAML) is a supervised learning framework that allows us to incorporate human expert knowledge into machine learning to build robust and dependable clinical predictive models. Problem-specific features are engineered using a state-of-the-art learning algorithm and a web application is used to collect human expert priors. The procedure allows us to discover problems in potentially noisy, biased clinical data with hidden confounders, identify shortcomings of human expert assessments, and build an optimized model with improved generalizability on unseen cases.
(Social Hour follwing seminar: 4:30-5PM)