Exact mean computation in dynamic time warping spaces

Averaging time series under dynamic time warping is an important tool for improving nearest-neighbor classifiers and formulating centroid-based clustering. The most promising approach poses time series averaging as the problem of minimizing a Fréchet function. Minimizing the Fréchet function is NP-hard and so far solved by several heuristics and inexact strategies. Our contributions are as follows: we first discuss some inaccuracies in the literature on exact mean computation in dynamic time warping spaces. Then we propose an exponential-time dynamic program for computing a global minimum of the Fréchet function. The proposed algorithm is useful for benchmarking and evaluating known heuristics. In addition, we present an exact polynomial-time algorithm for the special case of binary time series. Based on the proposed exponential-time dynamic program, we empirically study properties like uniqueness and length of a mean, which are of interest for devising better heuristics. Experimental evaluations indicate substantial deficits of state-of-the-art heuristics in terms of their output quality.

[1]  Geoffrey I. Webb,et al.  Dynamic Time Warping Averaging of Time Series Allows Faster and More Accurate Classification , 2014, 2014 IEEE International Conference on Data Mining.

[2]  C. Ratanamahatana,et al.  Shape averaging under Time Warping , 2009, 2009 6th International Conference on Electrical Engineering/Electronics, Computer, Telecommunications and Information Technology.

[3]  Marvin Künnemann,et al.  Quadratic Conditional Lower Bounds for String Problems and Dynamic Time Warping , 2015, 2015 IEEE 56th Annual Symposium on Foundations of Computer Science.

[4]  Ying Wah Teh,et al.  Time-series clustering - A decade review , 2015, Inf. Syst..

[5]  Pasi Fränti,et al.  Time-series clustering by approximate prototypes , 2008, ICPR.

[6]  Laurent Itti,et al.  shapeDTW: Shape Dynamic Time Warping , 2016, Pattern Recognit..

[7]  Vit Niennattrakul,et al.  Inaccuracies of Shape Averaging Method Using Dynamic Time Warping for Time Series Data , 2007, International Conference on Computational Science.

[8]  Eamonn J. Keogh,et al.  Derivative Dynamic Time Warping , 2001, SDM.

[9]  François Nicolas,et al.  Hardness results for the center and median string problems under the weighted and unweighted edit distances , 2005, J. Discrete Algorithms.

[10]  Pierre-François Marteau.  Times Series Averaging and Denoising from a Probabilistic Perspective on Time–Elastic Kernels , 2019, Int. J. Appl. Math. Comput. Sci..

[11]  Geoffrey I. Webb,et al.  Faster and more accurate classification of time series by exploiting a novel dynamic time warping averaging algorithm , 2015, Knowledge and Information Systems.

[12]  Tak-Chung Fu,et al.  A review on time series data mining , 2011, Eng. Appl. Artif. Intell..

[13]  Marco Cuturi,et al.  Soft-DTW: a Differentiable Loss Function for Time-Series , 2017, ICML.

[14]  R. Lummis,et al.  Speaker verification by computer using speech intensity for temporal registration , 1973 .

[15]  Laura Firoiu,et al.  Clustering Time Series with Hidden Markov Models and Dynamic Time Warping , 1999 .

[16]  George M. Church,et al.  Aligning gene expression time series with time warping algorithms , 2001, Bioinform..

[17]  David Schultz,et al.  On the Existence of a Sample Mean in Dynamic Time Warping Spaces , 2016 .

[18]  Pierre Gançarski,et al.  A global averaging method for dynamic time warping, with applications to clustering , 2011, Pattern Recognit..

[19]  L. Gupta,et al.  Nonlinear alignment and averaging for estimating the evoked potential , 1996, IEEE Transactions on Biomedical Engineering.

[20]  Sergio Escalera,et al.  Featureweighting in dynamic timewarping for gesture recognition in depth data , 2011, 2011 IEEE International Conference on Computer Vision Workshops (ICCV Workshops).

[21]  Jorge J. Moré,et al.  Digital Object Identifier (DOI) 10.1007/s101070100263 , 2001 .

[22]  Éric Gaussier,et al.  Generalized k-means-based clustering for temporal data under weighted and kernel time warp , 2016, Pattern Recognit. Lett..

[23]  Luis Gravano,et al.  Fast and Accurate Time-Series Clustering , 2017, ACM Trans. Database Syst..

[24]  Pierre Gançarski,et al.  Summarizing a set of time series by averaging: From Steiner sequence to compact multiple alignment , 2012, Theor. Comput. Sci..

[25]  Stan Sclaroff,et al.  A Unified Framework for Gesture Recognition and Spatiotemporal Gesture Segmentation , 2009, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[26]  T. Moon.  The expectation-maximization algorithm , 1996, IEEE Signal Process. Mag..

[27]  L. R. Rabiner,et al.  A comparative study of several dynamic time-warping algorithms for connected-word recognition , 1981, The Bell System Technical Journal.

[28]  Tomoko Matsui,et al.  A Kernel for Time Series Based on Global Alignments , 2006, 2007 IEEE International Conference on Acoustics, Speech and Signal Processing - ICASSP '07.

[29]  S. Chiba,et al.  Dynamic programming algorithm optimization for spoken word recognition , 1978 .

[30]  David Schultz,et al.  Optimal Warping Paths are unique for almost every pair of Time Series , 2017, ArXiv.

[31]  F. Clarke.  Optimization And Nonsmooth Analysis , 1983 .

[32]  Éric Gaussier,et al.  Progressive and Iterative Approaches for Time Series Averaging , 2015, AALTD@PKDD/ECML.

[33]  Micha Sharir,et al.  Dynamic Time Warping and Geometric Edit Distance: Breaking the Quadratic Barrier , 2016, ICALP.

[34]  Marimuthu Palaniswami,et al.  Fuzzy c-Shape: A new algorithm for clustering finite time series waveforms , 2016, 2017 IEEE International Conference on Fuzzy Systems (FUZZ-IEEE).

[35]  Paola Bonizzoni,et al.  The complexity of multiple sequence alignment with SP-score that is a metric , 2001, Theor. Comput. Sci..

[36]  Witold Kinsner,et al.  ECG frame classification using dynamic time warping , 2002, IEEE CCECE2002. Canadian Conference on Electrical and Computer Engineering. Conference Proceedings (Cat. No.02CH37373).

[37]  Rolf Niedermeier,et al.  Tight Hardness Results for Consensus Problems on Circular Strings and Time Series , 2018, SIAM J. Discret. Math..

[38]  C. L. Philip Chen,et al.  Degree-Pruning Dynamic Programming Approaches to Central Time Series Minimizing Dynamic Time Warping Distance , 2017, IEEE Transactions on Cybernetics.

[39]  M. Fréchet.  Les éléments aléatoires de nature quelconque dans un espace distancié , 1948 .

[40]  David Schultz,et al.  Nonsmooth analysis and subgradient methods for averaging in dynamic time warping spaces , 2017, Pattern Recognit..

[41]  Marion Morel,et al.  Time-series averaging using constrained dynamic time warping with tolerance , 2018, Pattern Recognit..

[42]  Jay G. Wilpon,et al.  Considerations in applying clustering techniques to speaker-independent word recognition. , 1979 .

[43]  김동규,et al.  [서평]「Algorithms on Strings, Trees, and Sequences」 , 2000 .

[44]  David Sankoff,et al.  Time Warps, String Edits, and Macromolecules: The Theory and Practice of Sequence Comparison , 1983 .

[45]  Waleed H. Abdulla,et al.  Cross-words reference template for DTW-based speech recognition systems , 2003, TENCON 2003. Conference on Convergent Technologies for Asia-Pacific Region.