Explore the cutting-edge field of conformal inference methods in this 35-minute lecture by Emmanuel Candès from Stanford University. Delve into the basic principles underlying conformal inference and discover how these methods provide exact prediction intervals for future observations without relying on distributional assumptions, requiring only iid or exchangeable data. Learn about recent major contributions in the field from the past 2-3 years, including enhanced conformity scores applicable to both quantitative and categorical labels. If time permits, gain insights into novel methods addressing situations where the distribution of observations can shift drastically. This talk, presented at the Institut des Hautes Etudes Scientifiques (IHES), offers a comprehensive overview of conformal prediction techniques that are gaining popularity in both academia and industry.
A Taste of Conformal Prediction - Principles and Recent Advances
Institut des Hautes Etudes Scientifiques (IHES) via YouTube
Overview
Syllabus
Emmanuel Candès - A Taste of Conformal Prediction
Taught by
Institut des Hautes Etudes Scientifiques (IHES)
Related Courses
-
Uncertainty Quantification in Machine Learning: Federated Conformal Prediction
-
Assumption-Free Prediction Intervals for Black-Box Regression Algorithms - Aaditya Ramdas
-
Uncertainty Quantification with Conformal Prediction - A Path to Reliable ML Models
-
Split Conformal Prediction and Its Extensions to Non-Exchangeable Data
-
Frugality for Sequential Prediction: Reaching Bounded Regret While Using Expert Advice Sparingly
-
UTOPIA - Universally Trainable Optimal Prediction Intervals Aggregation
