One Batch Is Enough: A Unified Dataset Condensation Framework for General Time Series Analysis

Wei Shao, Ziquan Fang, Zheqi Lu, Yongfeng Su, Yuzhu Wang, Yunjun Gao
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:110102-110130, 2026.

Abstract

Time-series analysis is critical in real-world applications, yet the explosion of time-series data imposes severe burdens on storage and computational resources. Recently, dataset condensation has emerged as a promising data-centric solution by synthesizing compact yet informative datasets to replace large-scale raw data. However, existing methods are largely vision-centric, failing to capture unique temporal properties of time series, or task-specific, tightly coupling the condensed data to a particular downstream objective. As a result, these approaches suffer from feature mismatch and fail to generalize across diverse time-series tasks. To bridge this gap, we propose UniTSC, the first unified dataset condensation framework for general time-series analysis. UniTSC employs a multi-view hybrid encoder to capture task-invariant representations across temporal, spectral, and topological perspectives. Building upon this representation, we design a tri-space alignment paradigm that jointly aligns optimization trajectories, power spectral densities, and multivariate dependency structures, enabling comprehensive information preservation under extreme compression. Extensive experiments show that UniTSC retains up to 97.9% of downstream performance using as little as 0.01% of the original training data, within our experimental settings tied to standard sequence lengths, revealing that a single batch-equivalent budget ($\textless$ 128 samples) is sufficient to capture the essential dynamics of complex time-series data.

Cite this Paper


BibTeX
@InProceedings{pmlr-v306-shao26b, title = {One Batch Is Enough: A Unified Dataset Condensation Framework for General Time Series Analysis}, author = {Shao, Wei and Fang, Ziquan and Lu, Zheqi and Su, Yongfeng and Wang, Yuzhu and Gao, Yunjun}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {110102--110130}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/shao26b/shao26b.pdf}, url = {https://proceedings.mlr.press/v306/shao26b.html}, abstract = {Time-series analysis is critical in real-world applications, yet the explosion of time-series data imposes severe burdens on storage and computational resources. Recently, dataset condensation has emerged as a promising data-centric solution by synthesizing compact yet informative datasets to replace large-scale raw data. However, existing methods are largely vision-centric, failing to capture unique temporal properties of time series, or task-specific, tightly coupling the condensed data to a particular downstream objective. As a result, these approaches suffer from feature mismatch and fail to generalize across diverse time-series tasks. To bridge this gap, we propose UniTSC, the first unified dataset condensation framework for general time-series analysis. UniTSC employs a multi-view hybrid encoder to capture task-invariant representations across temporal, spectral, and topological perspectives. Building upon this representation, we design a tri-space alignment paradigm that jointly aligns optimization trajectories, power spectral densities, and multivariate dependency structures, enabling comprehensive information preservation under extreme compression. Extensive experiments show that UniTSC retains up to 97.9% of downstream performance using as little as 0.01% of the original training data, within our experimental settings tied to standard sequence lengths, revealing that a single batch-equivalent budget ($\textless$ 128 samples) is sufficient to capture the essential dynamics of complex time-series data.} }
Endnote
%0 Conference Paper %T One Batch Is Enough: A Unified Dataset Condensation Framework for General Time Series Analysis %A Wei Shao %A Ziquan Fang %A Zheqi Lu %A Yongfeng Su %A Yuzhu Wang %A Yunjun Gao %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-shao26b %I PMLR %P 110102--110130 %U https://proceedings.mlr.press/v306/shao26b.html %V 306 %X Time-series analysis is critical in real-world applications, yet the explosion of time-series data imposes severe burdens on storage and computational resources. Recently, dataset condensation has emerged as a promising data-centric solution by synthesizing compact yet informative datasets to replace large-scale raw data. However, existing methods are largely vision-centric, failing to capture unique temporal properties of time series, or task-specific, tightly coupling the condensed data to a particular downstream objective. As a result, these approaches suffer from feature mismatch and fail to generalize across diverse time-series tasks. To bridge this gap, we propose UniTSC, the first unified dataset condensation framework for general time-series analysis. UniTSC employs a multi-view hybrid encoder to capture task-invariant representations across temporal, spectral, and topological perspectives. Building upon this representation, we design a tri-space alignment paradigm that jointly aligns optimization trajectories, power spectral densities, and multivariate dependency structures, enabling comprehensive information preservation under extreme compression. Extensive experiments show that UniTSC retains up to 97.9% of downstream performance using as little as 0.01% of the original training data, within our experimental settings tied to standard sequence lengths, revealing that a single batch-equivalent budget ($\textless$ 128 samples) is sufficient to capture the essential dynamics of complex time-series data.
APA
Shao, W., Fang, Z., Lu, Z., Su, Y., Wang, Y. & Gao, Y.. (2026). One Batch Is Enough: A Unified Dataset Condensation Framework for General Time Series Analysis. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:110102-110130 Available from https://proceedings.mlr.press/v306/shao26b.html.

Related Material