2022
DOI: 10.1109/tkde.2020.3033752
|View full text |Cite
|
Sign up to set email alerts
|

FastDTW is Approximate and Generally Slower Than the Algorithm it Approximates

Abstract: Many time series data mining problems can be solved with repeated use of distance measure. Examples of such tasks include similarity search, clustering, classification, anomaly detection and segmentation. For over two decades it has been known that the Dynamic Time Warping (DTW) distance measure is the best measure to use for most tasks, in most domains. Because the classic DTW algorithm has quadratic time complexity, many ideas have been introduced to reduce its amortized time, or to quickly approximate it. O… Show more

Help me understand this report
View preprint versions

Search citation statements

Order By: Relevance

Paper Sections

Select...
4
1

Citation Types

0
8
0

Year Published

2022
2022
2024
2024

Publication Types

Select...
5
3
2

Relationship

0
10

Authors

Journals

citations
Cited by 29 publications
(8 citation statements)
references
References 20 publications
0
8
0
Order By: Relevance
“…Modern speech recognition makes prevalent use of deep neural network models. Yet dynamic time warping has persisted as a useful method in data science [9,10] and cognitive science [11,12].…”
Section: Dynamic Time Warping As Absementmentioning
confidence: 99%
“…Modern speech recognition makes prevalent use of deep neural network models. Yet dynamic time warping has persisted as a useful method in data science [9,10] and cognitive science [11,12].…”
Section: Dynamic Time Warping As Absementmentioning
confidence: 99%
“…[28,41]) such as the Sakoe-Chiba Band and the Itakura parallelogram as these simplify the identification of the optimal warping path. While these are appealing concepts the authors in [42] observe that the wellknown FastDTW algorithm [41] is in fact slower than DTW.…”
Section: Dynamic Time Warpingmentioning
confidence: 99%
“…Pruning warping alignments is a valuable addition but has limited impact when many suitable warping paths exist. Approximate methods introduce an additional complexity that requires more memory and loses the computational simplicity of the original DTW algorithm making it often slow in practice [27]. In contrast, PQDTW is fast in its own right while maintaining compatibility with the aforementioned techniques.…”
Section: Dynamic Time Warping Dynamic Timementioning
confidence: 99%