Abstract
This book is about conformal prediction and related inferential techniquesthat build on permutation tests and exchangeability. These techniques areuseful in a diverse array of tasks, including hypothesis testing and providinguncertainty quantification guarantees for machine learning systems. Much of thecurrent interest in conformal prediction is due to its ability to integrateinto complex machine learning workflows, solving the problem of formingprediction sets without any assumptions on the form of the data generatingdistribution. Since contemporary machine learning algorithms have generallyproven difficult to analyze directly, conformal prediction's main appeal is itsability to provide formal, finite-sample guarantees when paired with suchmethods. The goal of this book is to teach the reader about the fundamental technicalarguments that arise when researching conformal prediction and relatedquestions in distribution-free inference. Many of these proof strategies,especially the more recent ones, are scattered among research papers, making itdifficult for researchers to understand where to look, which results areimportant, and how exactly the proofs work. We hope to bridge this gap bycurating what we believe to be some of the most important results in theliterature and presenting their proofs in a unified language, withillustrations, and with an eye towards pedagogy.