Asymmetric Learning Vector Quantization for Efficient Nearest Neighbor Classification in Dynamic Time Warping Spaces

TL;DR

This paper introduces an asymmetric learning vector quantization method tailored for dynamic time warping spaces, significantly improving efficiency and accuracy in large-scale time series nearest neighbor classification.

Contribution

It extends LVQ to DTW spaces using asymmetric weighted averaging, offering a novel approach for prototype generation in time series classification.

Findings

Asymmetric GLVQ outperforms existing prototype methods.

The method reduces storage and computation costs.

Empirical results show superior classification accuracy.

Abstract

The nearest neighbor method together with the dynamic time warping (DTW) distance is one of the most popular approaches in time series classification. This method suffers from high storage and computation requirements for large training sets. As a solution to both drawbacks, this article extends learning vector quantization (LVQ) from Euclidean spaces to DTW spaces. The proposed LVQ scheme uses asymmetric weighted averaging as update rule. Empirical results exhibited superior performance of asymmetric generalized LVQ (GLVQ) over other state-of-the-art prototype generation methods for nearest neighbor classification.