Faster Retrieval with a Two-Pass Dynamic-Time-Warping Lower Bound

TL;DR

This paper introduces a two-pass approach using a tighter lower bound for Dynamic Time Warping to significantly speed up time series similarity searches, achieving 2-3 times faster results.

Contribution

The paper proposes a novel two-pass method employing LB Improved, a tighter lower bound, to enhance the efficiency of DTW-based similarity search.

Findings

LB Improved is faster than LB Keogh for search tasks.

The approach achieves 2-3 times speedup on random-walk and shape time series.

Two-pass method reduces unnecessary DTW computations.

Abstract

The Dynamic Time Warping (DTW) is a popular similarity measure between time series. The DTW fails to satisfy the triangle inequality and its computation requires quadratic time. Hence, to find closest neighbors quickly, we use bounding techniques. We can avoid most DTW computations with an inexpensive lower bound (LB Keogh). We compare LB Keogh with a tighter lower bound (LB Improved). We find that LB Improved-based search is faster. As an example, our approach is 2-3 times faster over random-walk and shape time series.