aiqbench analysis
This research advances generative modeling for multivariate time series by applying flow matching in the wavelet domain to better capture multi-scale temporal structures and cross-channel dependencies. By leveraging discrete wavelet coefficients, the method implicitly handles coarse-to-fine generation while using a channel-token transformer to model inter-variable relationships without an explicit multi-scale schedule. The approach demonstrates consistent improvements over baselines across diverse benchmarks, offering a robust solution for data augmentation and privacy-preserving sharing in temporal domains.
Wavelet Flow Matching for Time Series
arXiv:2609.39374v1 Announce Type: cross Abstract: Synthetic time series are increasingly used for data augmentation, privacy-preserving data sharing, and downstream model development, yet faithfully reproducing both multi-scale temporal structure and cross-channel dependencies remains challenging. We study multivariate time-series generation through flow matching in the wavelet domain. By operating on multilevel discrete wavelet coefficients rather than directly in the time domain, the model represents coarse structure and progressively finer details at separate scales. Their naturally different variances further induce an implicit coarse-to-fine generative process without requiring an explicit multi-scale schedule. Since the transform acts independently on each channel, we pair it with a channel-token transformer whose attention directly models cross-channel dependencies. Across seven benchmark datasets and four sequence lengths, our method is best or tied on a majority of dataset-metric combinations, with the largest and most consistent improvements in Context-FID and discriminative score.
Abstract:Synthetic time series are increasingly used for data augmentation, privacy-preserving data sharing, and downstream model development, yet faithfully reproducing both multi-scale temporal structure and cross-channel dependencies remains challenging. We study multivariate time-series generation through flow matching in the wavelet domain. By operating on multilevel discrete wavelet coefficients rather than directly in the time domain, the model represents coarse structure and progressively finer details at separate scales. Their naturally different variances further induce an implicit coarse-to-fine generative process without requiring an explicit multi-scale schedule. Since the transform acts independently on each channel, we pair it with a channel-token transformer whose attention directly models cross-channel dependencies. Across seven benchmark datasets and four sequence lengths, our method is best or tied on a majority of dataset-metric combinations, with the largest and most consistent improvements in Context-FID and discriminative score.
Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
arXiv:2609.39374 [cs.LG]
(or arXiv:2609.39374v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2609.39374 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 233 words
Related coverage
Time series generation with spectrally aligned latent flow matching
arXiv · Sep 21, 2026
MGD: Moment Guided Diffusion for Maximum Entropy Generation
arXiv · Sep 9, 2026
MercerFlow: Flow Matching in a Kernel-Induced Latent Space for Probabilistic Forecasting
arXiv · Oct 7, 2026
Graph-Spectral Flow Matching for Multivariate Time Series Anomaly Detection
arXiv · Sep 30, 2026