Problem
Research questions and friction points this paper is trying to address.
synthetic time series
multivariate time-series generation
multi-scale temporal structure
cross-channel dependencies
Wavelet Flow Matching for Time Series
🤖 AI Summary
This study addresses the challenge of multivariate time series generation, where simultaneously reproducing multi-scale temporal structures and cross-channel dependencies remains difficult. To this end, we propose a generative framework that applies flow matching within the discrete wavelet coefficient domain. By leveraging the inherent variance disparities of the wavelet transform, our approach induces an implicit coarse-to-fine generation mechanism without requiring explicit scheduling, thereby effectively modeling multi-scale structures. Furthermore, a channel-token-based Transformer is integrated to capture cross-channel dependencies, enabling the joint modeling of both aspects. Extensive evaluations across seven benchmark datasets demonstrate that the proposed method achieves leading performance on most metrics, yielding significant and consistent improvements in Context-FID and discriminative scores.
📝 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.
Innovation
Methods, ideas, or system contributions that make the work stand out.
Wavelet Flow Matching
Multivariate Time Series Generation
Coarse-to-Fine Generation
Channel-Token Transformer
Cross-Channel Dependencies
🔎 Similar Papers
No similar papers found.
