FracTS: Hierarchical and Autoregressive Time Series Generation

Key points

• Time series are ubiquitous across different domains, and the growing use of synthetic data for privacy-preserving sharing, data augmentation, and stress testing makes time series generation increasingly important.1
• Early generative approaches are predominantly autoregressive, which can suffer from low sample diversity, error accumulation, and difficulty modeling long-range dependencies.1
• Inspired by fractal generative modeling developed for images, FracTS adapts the paradigm from 2D to 1D sequences by learning a multi-scale representation that captures global trajectory structure at coarse scales and local dynamics at fine scales while following the autoregressive nature of time series.1
• Experiments on five datasets (two conditional, three unconditional) conducted against seven baselines show that FracTS consistently outperforms baselines across marginal distribution and predictive ML utility, and correlation preservation in the unconditional setting.1

Sources (1)

• [1]FracTS: Hierarchical and Autoregressive Time Series Generation
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 18, 08:35 PM

  In this paper, we introduce FracTS, a hierarchical, coarse-to-fine autoregressive model for multivariate time series generation.

  Time series are ubiquitous across different domains, and the growing use of synthetic data for privacy-preserving sharing, data augmentation, and stress testing makes time series generation increasingly important.

Extractive summary: sentences quoted from the sources.