Abstract
Reducing the Impact of Label Noise in AI Applications in Healthcare
Dr. Marzyeh Ghassemi, Pilot Investigator, Associate Professor at MIT in Electrical Engineering and Computer Science (EECS) and Institute for Medical Engineering & Science (IMES)
The learning that takes place in artificial intelligence applications uses data that may be incomplete, imprecise, distorted, or mislabeled, thereby tainting the interpretations made by the AI algorithm. The goal of this project is to benchmark the impact of such label noise in at least four datasets, including a clinical dataset; and then test and evaluate a method to reduce its distortionary impact.