Closing the Loop: Co-Evolving EM for Irregular Time Series Generation in Lifted Representations
Paper · https://arxiv.org/abs/2611.14659
arXiv:2611.14659v1 Announce Type: new Abstract: Regular time-series generation from irregular time series is typically reduced to one-shot imputation: complete each partial sequence once, freeze the completions, and train a generator on the resulting surrogate dataset. We show that this open-loop strategy is structurally flawed. Point imputers collapse ambiguous missing regions to conditional averages, while stochastic imputers become stale as the generator evolves; the binding constraint is therefore iteration, not imputer design. We introduce a co-evolving Monte Carlo EM framework that closes the imputer-generator loop. Each E-step samples missing values from the posterior of the current generator, and each M-step retrains the generator on these completions, preserving unconditional generator training. Since our diffusion prior operates in a lifted 2D representation while observations live in time-series space, we further introduce Posterior Sampling for Lifted Representations (PSLR), which enforces operator-consistent conditioning, uncertainty preservation, and manifold consistency during the E-step. Across ten datasets and four corruption regimes, our method achieves state-of-the-art performance, improving over the strongest baselines by 59% in discriminative score, 15% in predictive score, 73% in Context-FID, and 71% in feature-correlation error. Ablations show that our approach reduces average discriminative score by 73% over single-step baselines, closing most of the remaining gap to the clean-data oracle. Code is available at https://github.com/azencot-group/PSLR.
First seen by Oriel on .
AI summary
The paper presents a co-evolving Monte Carlo EM framework for generating regular time series from irregular observations, substantially outperforming one-shot imputation baselines across ten datasets and four corruption regimes.
- Each E-step samples missing values from the current generator’s posterior, while each M-step retrains the generator on those completions without changing unconditional training.
- Posterior Sampling for Lifted Representations (PSLR) connects time-series observations to a lifted 2D diffusion prior while preserving uncertainty, operator-consistent conditioning, and manifold consistency.
- The method improves over the strongest baselines by 59% in discriminative score, 15% in predictive score, 73% in Context-FID, and 71% in feature-correlation error.
- Ablations reduce average discriminative score by 73% versus single-step baselines, closing most of the gap to the clean-data oracle.
- The authors argue iteration is the key constraint because point imputers average ambiguous regions and stochastic imputations become stale as the generator evolves.
#time series #monte carlo em #diffusion models #missing data
Generated by gpt-5.6-sol. Check the original for details.
Aliases
| Source | External ID | URL |
|---|---|---|
| arXiv AI Papers | oai:arXiv.org:2611.14659v1 | https://arxiv.org/abs/2611.14659 Read discussion |