21
1

A Two-Stage Active Learning Algorithm for kk-Nearest Neighbors

Abstract

kk-nearest neighbor classification is a popular non-parametric method because of desirable properties like automatic adaption to distributional scale changes. Unfortunately, it has thus far proved difficult to design active learning strategies for the training of local voting-based classifiers that naturally retain these desirable properties, and hence active learning strategies for kk-nearest neighbor classification have been conspicuously missing from the literature. In this work, we introduce a simple and intuitive active learning algorithm for the training of kk-nearest neighbor classifiers, the first in the literature which retains the concept of the kk-nearest neighbor vote at prediction time. We provide consistency guarantees for a modified kk-nearest neighbors classifier trained on samples acquired via our scheme, and show that when the conditional probability function P(Y=yX=x)\mathbb{P}(Y=y|X=x) is sufficiently smooth and the Tsybakov noise condition holds, our actively trained classifiers converge to the Bayes optimal classifier at a faster asymptotic rate than passively trained kk-nearest neighbor classifiers.

View on arXiv
Comments on this paper