Time series generation with spectrally aligned latent flow matching
Spectrally aligned latent flow matching generates time series that preserve Fourier, wavelet, and signature structure.
Latent flow models compress time series but can distort spectra, limiting their use as training surrogates. The authors train the latent space for flow matching with fine-tuning losses from Fourier, wavelet, and signature transforms so synthetic samples keep relevant dynamics. They compare the aligned generator with a base latent-flow model and prior methods on long-range univariate and multivariate benchmarks. Reported gains cover signal realism and compute cost while matching the local structure of the training data.
- Spectral mismatch is a known artefact of latent time-series generators.
- Fine-tuning uses Fourier, wavelet, and signature-transform losses.
- Models are evaluated on long-range univariate and multivariate benchmarks.
- Authors report better realism and efficiency than a base latent-flow model.
Full article198 words · extracted from arxiv.org · click to collapse
Latent flow models have proven to be a reliable and cost-effective method for time series generation. However, the latent compression induces unwanted artefacts, such as a spectral mismatch with respect to the underlying dataset, thus hindering their use as training surrogates. In this article, we propose a spectrally-aligned latent-flow time series generator, where the latent space for flow matching is trained to preserve dynamical properties that are relevant for the suitability of synthetic samples. We find that incorporating fine-tuning losses based on canonical signal representations such as the Fourier, wavelet and signature transforms helps overcome these issues. The interpretability of these transformations allows us to ensure that the synthetic signals are aligned with the true ones in terms of relevant features, such as smoothness or targeted spectral content, as opposed to relying on pointwise reconstruction losses only. We compare the proposed aligned models against a base latent-flow model and the state of the art over real-world long-range univariate and multivariate benchmark datasets. Our quantitative results validate the superiority of the proposed method in terms of its performance on metrics reflecting signal realness and computational efficiency, while being aligned to the training set with respect to its local structure.
Text extracted automatically; images, tables and formatting may be missing. Original: https://arxiv.org/abs/2609.21989