Textbook of AI

Abstract. Introduces the lasso (L1-penalised regression), which simultaneously performs feature selection and coefficient estimation by driving some weights exactly to zero. The paper established sparsity-inducing regularisation as a fundamental tool in statistics and machine learning.