Algorithmic Learning in a Random World

Conformal prediction is a machine learning approach that treats prediction reliability and is valid, flexible, and efficient. This Second Edition covers key conformal predictors, mathematical analysis, and applications in medicine and industry. It also includes new chapters on conformal predictive distributions and testing procedures for protecting algorithms against data distribution changes.

Format: Hardback
Length: 476 pages
Publication date: 14 December 2022
Publisher: Springer International Publishing AG

Conformal prediction is a machine learning approach that emerged in the late 1990s, focusing on the reliability of predictions. It involves the use of prediction algorithms known as conformal predictors, which are designed to evaluate the accuracy of their own predictions in a balanced and unbiased manner. These algorithms are proven to be valid, meaning they avoid being overly pessimistic or overly optimistic.

The flexibility of conformal prediction allows it to incorporate most existing powerful methods of machine learning. The book covers both key conformal predictors and the mathematical analysis of their properties.

In addition to proofs of validity, the book provides results about the efficiency of conformal predictors. The assumption of randomness, which is fundamental to the approach, is relaxed in later chapters, allowing for more practical applications.

Since its publication in 2005, conformal prediction has found numerous applications in medicine and industry, and is becoming a popular machine-learning technique. The Second Edition of this book includes three new chapters. One chapter focuses on conformal predictive distributions, which offer more informative predictions than standard conformal predictors. Another chapter discusses the efficiency of testing the assumption of randomness based on conformal prediction. The third chapter harnesses conformal testing procedures to protect machine-learning algorithms against changes in the data.

Overall, Conformal Prediction: An Approach to Prediction with Reliable Predictions is a valuable resource for researchers and practitioners in the field of machine learning, providing a comprehensive understanding of this important approach to prediction.

Weight: 916g
Dimension: 235 x 155 (mm)
ISBN-13: 9783031066481
Edition number: 2nd ed. 2022