Coarse-DTW for sparse time series alignment

Abstract

Dynamic Time Warping (DTW) is considered as a robust measure to compare numerical time series when some time elasticity is required. However, speed is a known major drawback of DTW due to its quadratic complexity. Previous work has mainly considered designing speed optimization based on early-abandoning strategies applied to nearest-neighbor classification, although some of these optimizations are restricted to uni-dimensional time series. In this paper, we introduce Coarse-DTW, a reinterpretation of DTW for sparse time series, which exploits adaptive downsampling to achieve speed enhancement, even when faced with multidimensional time series. We show that Coarse-DTW achieves nontrivial speedups in nearest-neighbor classification and even admits a positive-definite kernelization suitable for SVM classification, hence offering a good tradeoff between speed and accuracy.