TimePoint: Accelerated Time Series Alignment via Self-Supervised Keypoint and Descriptor Learning

Ron Shapira Weber, Shahar Benishay, Andrey Lavrinenko, Shahaf E. Finder, Oren Freifeld
Proceedings of the 42nd International Conference on Machine Learning, PMLR 267:54275-54299, 2025.

Abstract

Fast and scalable alignment of time series is a fundamental challenge in many domains. The standard solution, Dynamic Time Warping (DTW), struggles with poor scalability and sensitivity to noise. We introduce TimePoint, a self-supervised method that dramatically accelerates DTW-based alignment while typically improving alignment accuracy by learning keypoints and descriptors from synthetic data. Inspired by 2D keypoint detection but carefully adapted to the unique challenges of 1D signals, TimePoint leverages efficient 1D diffeomorphisms, which effectively model nonlinear time warping, to generate realistic training data. This adaptation, along with fully convolutional and wavelet convolutional architectures, enables the extraction of informative keypoints and descriptors. Applying DTW to these sparse representations yields major speedups and typically higher alignment accuracy than standard DTW applied to the full signals. Despite being trained solely on synthetic data, TimePoint generalizes well to real-world time series. Extensive experiments demonstrate that TimePoint consistently achieves faster and more accurate alignments than standard DTW, making it a scalable solution for time-series analysis. Our code is available at https://github.com/ BGU-CS-VIL/TimePoint.

Cite this Paper

BibTeX
@InProceedings{pmlr-v267-shapira-weber25a, title = {{T}ime{P}oint: Accelerated Time Series Alignment via Self-Supervised Keypoint and Descriptor Learning}, author = {Shapira Weber, Ron and Benishay, Shahar and Lavrinenko, Andrey and Finder, Shahaf E. and Freifeld, Oren}, booktitle = {Proceedings of the 42nd International Conference on Machine Learning}, pages = {54275--54299}, year = {2025}, editor = {Singh, Aarti and Fazel, Maryam and Hsu, Daniel and Lacoste-Julien, Simon and Berkenkamp, Felix and Maharaj, Tegan and Wagstaff, Kiri and Zhu, Jerry}, volume = {267}, series = {Proceedings of Machine Learning Research}, month = {13--19 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v267/main/assets/shapira-weber25a/shapira-weber25a.pdf}, url = {https://proceedings.mlr.press/v267/shapira-weber25a.html}, abstract = {Fast and scalable alignment of time series is a fundamental challenge in many domains. The standard solution, Dynamic Time Warping (DTW), struggles with poor scalability and sensitivity to noise. We introduce TimePoint, a self-supervised method that dramatically accelerates DTW-based alignment while typically improving alignment accuracy by learning keypoints and descriptors from synthetic data. Inspired by 2D keypoint detection but carefully adapted to the unique challenges of 1D signals, TimePoint leverages efficient 1D diffeomorphisms, which effectively model nonlinear time warping, to generate realistic training data. This adaptation, along with fully convolutional and wavelet convolutional architectures, enables the extraction of informative keypoints and descriptors. Applying DTW to these sparse representations yields major speedups and typically higher alignment accuracy than standard DTW applied to the full signals. Despite being trained solely on synthetic data, TimePoint generalizes well to real-world time series. Extensive experiments demonstrate that TimePoint consistently achieves faster and more accurate alignments than standard DTW, making it a scalable solution for time-series analysis. Our code is available at https://github.com/ BGU-CS-VIL/TimePoint.} }
Endnote
%0 Conference Paper %T TimePoint: Accelerated Time Series Alignment via Self-Supervised Keypoint and Descriptor Learning %A Ron Shapira Weber %A Shahar Benishay %A Andrey Lavrinenko %A Shahaf E. Finder %A Oren Freifeld %B Proceedings of the 42nd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2025 %E Aarti Singh %E Maryam Fazel %E Daniel Hsu %E Simon Lacoste-Julien %E Felix Berkenkamp %E Tegan Maharaj %E Kiri Wagstaff %E Jerry Zhu %F pmlr-v267-shapira-weber25a %I PMLR %P 54275--54299 %U https://proceedings.mlr.press/v267/shapira-weber25a.html %V 267 %X Fast and scalable alignment of time series is a fundamental challenge in many domains. The standard solution, Dynamic Time Warping (DTW), struggles with poor scalability and sensitivity to noise. We introduce TimePoint, a self-supervised method that dramatically accelerates DTW-based alignment while typically improving alignment accuracy by learning keypoints and descriptors from synthetic data. Inspired by 2D keypoint detection but carefully adapted to the unique challenges of 1D signals, TimePoint leverages efficient 1D diffeomorphisms, which effectively model nonlinear time warping, to generate realistic training data. This adaptation, along with fully convolutional and wavelet convolutional architectures, enables the extraction of informative keypoints and descriptors. Applying DTW to these sparse representations yields major speedups and typically higher alignment accuracy than standard DTW applied to the full signals. Despite being trained solely on synthetic data, TimePoint generalizes well to real-world time series. Extensive experiments demonstrate that TimePoint consistently achieves faster and more accurate alignments than standard DTW, making it a scalable solution for time-series analysis. Our code is available at https://github.com/ BGU-CS-VIL/TimePoint.
APA
Shapira Weber, R., Benishay, S., Lavrinenko, A., Finder, S.E. & Freifeld, O.. (2025). TimePoint: Accelerated Time Series Alignment via Self-Supervised Keypoint and Descriptor Learning. Proceedings of the 42nd International Conference on Machine Learning, in Proceedings of Machine Learning Research 267:54275-54299 Available from https://proceedings.mlr.press/v267/shapira-weber25a.html.

Related Material