Recent Advances in Conformal Prediction

• Activity Number: 278 - Emerging Ideas in Predictive Inference
• Type: Invited
• Date/Time: Tuesday, July 30, 2019 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistical Learning and Data Science
• Author(s): Larry Wasserman*
• Companies: Carnegie Mellon University
• Keywords: prediction; distribution-free; exchangeability

Abstract

Conformal prediction is a method for distribution-free prediction that assumes only exchangeability. In this talk, I'll review the conformal prediction and how it can be applied to a variety of problems. I will also explain how some problems, such as clustering, can be re-cast as prediction problems. Then I will discuss the problem of detecting lack of exchangeability and how to deal with it.