EECS Special Seminar: Stephen Bates, “A Statistical Toolbox for Deep Learning Models: Using ML Systems for Science”
Broad Institute Auditorium
Using learning algorithms to make consequential decisions in science and medicine requires a precise understanding of the uncertainty of the algorithms’ outputs. In this talk, we introduce tools for rigorous statistical inference with machine learning systems. The techniques yield confidence intervals, p-values, and other notions of statistical error control that are valid with no assumptions on the (unknown) data-generating distribution or the machine learning model. This allows researchers to leverage sophisticated machine-learning systems to draw reliable conclusions from data. We demonstrate the proposed techniques in imaging and proteomics applications.
- Date: Monday, March 20
- Time: 3:00 pm - 4:00 pm
- Location: Broad Institute Auditorium