Metric learning for temporal sequence alignment
2014
Conference Paper
In this paper, we propose to learn a Mahalanobis distance to perform alignment of multivariate time series. The learning examples for this task are time series for which the true alignment is known. We cast the alignment problem as a structured prediction task, and propose realistic losses between alignments for which the optimization is tractable. We provide experiments on real data in the audio-toaudio context, where we show that the learning of a similarity measure leads to improvements in the performance of the alignment task. We also propose to use this metric learning framework to perform feature selection and, from basic audio features, build a combination of these with better alignment performance.
Author(s): | Garreau, Damien and Lajugie, Rémi and Arlot, Sylvain and Bach, Francis |
Book Title: | Advances in Neural Information Processing Systems |
Pages: | 1817--1825 |
Year: | 2014 |
Bibtex Type: | Conference Paper (inproceedings) |
URL: | https://arxiv.org/abs/1409.3136 |
Links: |
Paper
|
BibTex @inproceedings{garreau2014metric, title = {Metric learning for temporal sequence alignment}, author = {Garreau, Damien and Lajugie, R{\'e}mi and Arlot, Sylvain and Bach, Francis}, booktitle = {Advances in Neural Information Processing Systems}, pages = {1817--1825}, year = {2014}, doi = {}, url = {https://arxiv.org/abs/1409.3136} } |