Linear Discrimination with Adaptive Ridge Classification Rules

This article considers the use of adaptive ridge classification rules for classifying an observation as coming from one of two multivariate normal distributionsN(μ⁽¹⁾, Σ) andN(μ⁽²⁾, Σ). In particular, the asymptotic expected error rates for a general class of these rules are obtained and are compared with that of the usual linear discriminant rule.