Voice2Series: Reprogramming Acoustic Models for Time Series Classification

Chao-Han Huck Yang, Yun-Yun Tsai, Pin-Yu Chen
Proceedings of the 38th International Conference on Machine Learning, PMLR 139:11808-11819, 2021.

Abstract

Learning to classify time series with limited data is a practical yet challenging problem. Current methods are primarily based on hand-designed feature extraction rules or domain-specific data augmentation. Motivated by the advances in deep speech processing models and the fact that voice data are univariate temporal signals, in this paper we propose Voice2Serie (V2S), a novel end-to-end approach that reprograms acoustic models for time series classification, through input transformation learning and output label mapping. Leveraging the representation learning power of a large-scale pre-trained speech processing model, on 31 different time series tasks we show that V2S outperforms or is on part with state-of-the-art methods on 22 tasks, and improves their average accuracy by 1.72%. We further provide theoretical justification of V2S by proving its population risk is upper bounded by the source risk and a Wasserstein distance accounting for feature alignment via reprogramming. Our results offer new and effective means to time series classification.

Cite this Paper


BibTeX
@InProceedings{pmlr-v139-yang21j, title = {Voice2Series: Reprogramming Acoustic Models for Time Series Classification}, author = {Yang, Chao-Han Huck and Tsai, Yun-Yun and Chen, Pin-Yu}, booktitle = {Proceedings of the 38th International Conference on Machine Learning}, pages = {11808--11819}, year = {2021}, editor = {Meila, Marina and Zhang, Tong}, volume = {139}, series = {Proceedings of Machine Learning Research}, month = {18--24 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v139/yang21j/yang21j.pdf}, url = {https://proceedings.mlr.press/v139/yang21j.html}, abstract = {Learning to classify time series with limited data is a practical yet challenging problem. Current methods are primarily based on hand-designed feature extraction rules or domain-specific data augmentation. Motivated by the advances in deep speech processing models and the fact that voice data are univariate temporal signals, in this paper we propose Voice2Serie (V2S), a novel end-to-end approach that reprograms acoustic models for time series classification, through input transformation learning and output label mapping. Leveraging the representation learning power of a large-scale pre-trained speech processing model, on 31 different time series tasks we show that V2S outperforms or is on part with state-of-the-art methods on 22 tasks, and improves their average accuracy by 1.72%. We further provide theoretical justification of V2S by proving its population risk is upper bounded by the source risk and a Wasserstein distance accounting for feature alignment via reprogramming. Our results offer new and effective means to time series classification.} }
Endnote
%0 Conference Paper %T Voice2Series: Reprogramming Acoustic Models for Time Series Classification %A Chao-Han Huck Yang %A Yun-Yun Tsai %A Pin-Yu Chen %B Proceedings of the 38th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2021 %E Marina Meila %E Tong Zhang %F pmlr-v139-yang21j %I PMLR %P 11808--11819 %U https://proceedings.mlr.press/v139/yang21j.html %V 139 %X Learning to classify time series with limited data is a practical yet challenging problem. Current methods are primarily based on hand-designed feature extraction rules or domain-specific data augmentation. Motivated by the advances in deep speech processing models and the fact that voice data are univariate temporal signals, in this paper we propose Voice2Serie (V2S), a novel end-to-end approach that reprograms acoustic models for time series classification, through input transformation learning and output label mapping. Leveraging the representation learning power of a large-scale pre-trained speech processing model, on 31 different time series tasks we show that V2S outperforms or is on part with state-of-the-art methods on 22 tasks, and improves their average accuracy by 1.72%. We further provide theoretical justification of V2S by proving its population risk is upper bounded by the source risk and a Wasserstein distance accounting for feature alignment via reprogramming. Our results offer new and effective means to time series classification.
APA
Yang, C.H., Tsai, Y. & Chen, P.. (2021). Voice2Series: Reprogramming Acoustic Models for Time Series Classification. Proceedings of the 38th International Conference on Machine Learning, in Proceedings of Machine Learning Research 139:11808-11819 Available from https://proceedings.mlr.press/v139/yang21j.html.

Related Material