Ridge Regression: A Historical Context

Abstract

Two classical articles on Ridge Regression by Arthur Hoerl and Robert Kennard were published in Technometrics in 1970, making 2020 their 50th anniversary. The theory and practice of Ridge Regression, and of related biased shrinkage estimators, have been extensively developed over the years. Further, newer shrinkage estimators, such as the Lasso and the Elastic Net, have become popular more recently. These newer developments have led to renewed interest in the original 1970 articles. What has perhaps been lost since 1970 is the context of these classic articles. That is, who were Art Hoerl and Bob Kennard, and what led two statisticians working in the private sector to develop Ridge Regression in the first place? What are the origins of Ridge Regression? Where did the name come from? The purpose of this article is to provide this historical context by discussing the men involved, their work at DuPont, and their approach to methodological development. As Art Hoerl was my father, this is admittedly a personal viewpoint.