New Paradigms in Modern Statistics I

Intro

Why distribution-free inference matters once models become black boxes, and the primary sources this volume draws on.

In classical statistics, how we treat prediction depends on strong parametric assumptions. With the incredible growth of AI, where the model is effectively a black box, distribution-free inference is becoming a substantial field in statistics. And it is not only about inference: controlling the risk of predictions and detecting out-of-distribution (OOD) inputs are critical to controlling the quality of these machines.

Here, I introduce the most generic and simple framework among them, conformal prediction, and review its representative papers. The primary sources I reference in this section are listed below.

Primary sources

Anastasios N. Angelopoulos. A Tutorial on Conformal Prediction. YouTube.

Anastasios N. Angelopoulos, Rina Foygel Barber, Stephen Bates. Theoretical Foundations of Conformal Prediction. arXiv:2411.11824, 2024.

Anastasios N. Angelopoulos, Stephen Bates, Emmanuel J. Candès, Michael I. Jordan, Lihua Lei. Learn then Test: Calibrating Predictive Algorithms to Achieve Risk Control. The Annals of Applied Statistics 19(2), 2025.

Anastasios N. Angelopoulos, Stephen Bates. A Gentle Introduction to Conformal Prediction and Distribution-Free Uncertainty Quantification. arXiv:2107.07511, 2021.