Diffusion-based Time Series Imputation and Forecasting with Structured State Space Models

By Juan Miguel Lopez Alcaraz, Nils Strodthoff. Trans. Mach. Learn. Res., 2022

302 citations recorded by the source corpus.

The imputation of missing values represents a significant obstacle for many real-world data analysis pipelines. Here, we focus on time series data and put forward SSSD, an imputation model that relies on two emerging technologies, (conditional) diffusion models as state-of-the-art generative models and structured state space models as internal model architecture, which are particularly suited to capture long-term dependencies in time series data. We demonstrate that SSSD matches or even exceeds state-of-the-art probabilistic imputation and forecasting performance on a broad range of data sets and different missingness scenarios, including the challenging blackout-missing scenarios, where prior approaches failed to provide meaningful results.

View Paper: https://www.semanticscholar.org/paper/ef7993ab30d0a8afabb4ebab080e471c0d5c743c

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

1. Ling Yang
2. Runsheng Xu
3. Zhangyang Gao
4. P. Heng
5. Cheng Tan
6. Ming-Hsuan Yang
7. Zhilong Zhang
8. Yue Zhao

Citations in the corpus

• Diffusion Models: A Comprehensive Survey of Methods and Applications (2022, methodology): “Structured State Space Diffusion (SSSD) [1] integrates conditional diffusion models and structured state-space models [82] to particularly capture long-term dependencies in time series.”
• A Survey on Generative Diffusion Models (2022, methodology): “In inpainting tasks, observed states are used as prior conditions [17], [180], [181], [182], combined with context-based modules.”