Dynamic Time Warping under limited warping path length

被引:62
作者
Zhang, Zheng [1 ]
Tavenard, Romain [2 ]
Bailly, Adeline [2 ]
Tang, Xiaotong [3 ]
Tang, Ping [1 ]
Corpetti, Thomas [2 ,4 ]
机构
[1] Chinese Acad Sci, Inst Remote Sensing & Digital Earth RADI, 20 Datun Rd, Beijing 100101, Peoples R China
[2] Univ Rennes 2, COSTEL, LETG Rennes, UMR 6554, F-35043 Rennes, France
[3] Northeastern Univ, Qinhuangdao, Hebei, Peoples R China
[4] CNRS, LETG Rennes, COSTEL UMR 6554, Pl Recteur Henri Moal, F-35043 Rennes, France
关键词
Dynamic Time Warping (DTW); Time series; Distance measure; Warping path; Classification; PROGRAMMING ALGORITHM; SERIES DATA;
D O I
10.1016/j.ins.2017.02.018
中图分类号
TP [自动化技术、计算机技术];
学科分类号
0812 ;
摘要
Dynamic Time Warping (DTW) is probably the most popular distance measure for time series data, because it captures flexible similarities under time distortions. However, DTW has long been suffering from the pathological alignment problem, and most existing solutions, which essentially impose rigid constraints on the warping path, are likely to miss the correct alignments. A crucial observation on pathological alignment is that it always leads to an abnormally large number of links between two sequences. Based on this new observation, we propose a novel variant of DTW called LDTW, which limits the total number of links during the optimization process of DTW. LDTW not only oppresses the pathological alignment effectively, but also allows more flexibilities when measuring similarities. It is a softer constraint because we still let the optimization process of DTW decide how many links to allocate to each data point and where to put these links. In this paper, we introduce the motivation and algorithm of LDTW and we conduct a nearest neighbor classification experiment on UCR time series archive to show its performance. (C) 2017 Elsevier Inc. All rights reserved.
引用
收藏
页码:91 / 107
页数:17
相关论文
共 37 条
[1]  
Bailly A., 2015, ECML PKDD WORKSH ADV
[2]   WARP: Accurate retrieval of shapes using phase of Fourier descriptors and time warping distance [J].
Bartolini, I ;
Ciaccia, P ;
Patella, M .
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, 2005, 27 (01) :142-147
[3]  
Batista G. E., 2011, P 2011 SIAM INT C DA, P699, DOI DOI 10.1137/1.9781611972818.60
[4]   Shape matching and object recognition using shape contexts [J].
Belongie, S ;
Malik, J ;
Puzicha, J .
IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE, 2002, 24 (04) :509-522
[5]  
Bergmann B., 1988, Multiple Hypotheses Testing, P100
[6]  
Berndt D.J., 1994, P 3 INT C KNOWLEDGE, P359, DOI DOI 10.5555/3000850.3000887
[7]   sDTW: Computing DTW Distances using Locally Relevant Constraints based on Salient Feature Alignments [J].
Candan, K. Selcuk ;
Rossini, Rosaria ;
Sapino, Maria Luisa ;
Wang, Xiaolan .
PROCEEDINGS OF THE VLDB ENDOWMENT, 2012, 5 (11) :1519-1530
[8]  
Chen YP, 2013, 19TH ACM SIGKDD INTERNATIONAL CONFERENCE ON KNOWLEDGE DISCOVERY AND DATA MINING (KDD'13), P383
[9]  
Cormen T. H., 2009, Introduction to Algorithms
[10]  
Ding H, 2008, PROC VLDB ENDOW, V1, P1542