Kick-one-out-based variable selection method for Euclidean distance-based classifier in high-dimensional settings (Q2034468)

The Euclidean distance-based classification in high-dimensional feature spaces is one of the core methods in machine learning applications. The authors analyse the influence of redundant variables dimension on the expected probabilities of misclassification. The main result of this paper is the consistency proof of the variable selection method such as kick-one-out based criteria, in the context of high dimensionality.