Research map / Time series / Causal and generative time series
TEDM: Time Series Forecasting with Elucidated Diffusion Models
Edgardo Solano Carrillo, Sreerag V Naveenachandran, Julia Niebling
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| replace preset EDM schedules with empirical data-driven noise and scale schedules (cumulative) (ablation) | EDM · ETTm2 (H=96) | MSE: 0.293 -> 0.135 | Improved |
| replace preset EDM schedules with empirical data-driven noise and scale schedules (cumulative) (ablation) | EDM · ETTh2 (H=96) | MSE: 0.419 -> 0.242 | Improved |
| replace preset EDM schedules with empirical data-driven noise and scale schedules (cumulative) (ablation) | EDM · Exchange (H=96) | MSE: 0.448 -> 0.068 | Improved |
| use empirical scale st instead of st = 1 (ablation) | TEDM (cumulative) · Exchange (H=96) | MSE: 0.110 -> 0.068 | Improved |
| use empirical scale st instead of st = 1 (ablation) | TEDM (cumulative) · ETTh2 (H=96) | MSE: 0.303 -> 0.242 | Improved |
| use empirical scale st instead of st = 1 (ablation) | TEDM (cumulative) · ETTm2 (H=96) | MSE: 0.137 -> 0.135 | No clear effect |
| use sliding window estimation instead of cumulative estimation (ablation) | TEDM · ETTh2 (H=96) | MSE: 0.242 -> 0.216 | Improved |
| use sliding window estimation instead of cumulative estimation (ablation) | TEDM · ETTm2 (H=96) | MSE: 0.135 -> 0.142 | Hurt |
| use sliding window estimation instead of cumulative estimation (ablation) | TEDM · Exchange (H=96) | MSE: 0.068 -> 0.075 | Hurt |
| use EDM instead of iDDPM+DDIM (ablation) | iDDPM+DDIM · ETTm2 (H=96) | MSE: 0.756 -> 0.293 | Improved |
| use EDM instead of iDDPM+DDIM (ablation) | iDDPM+DDIM · Exchange (H=96) | MSE: 1.276 -> 0.448 | Improved |
| align diffusion time axis with physical time (one Euler step per forecast step) | TEDM vs diffusion baselines · ETTm2 training time | Train Time (s): 0.107 -> 0.004 | Improved |
Closest papers by meaning
- Utilizing Image Transforms and Diffusion Models for Generative Modeling of Short and Long Time Series (NeurIPS 2024 · similarity 0.90)
- A Diffusion Model for Regular Time Series Generation from Irregular Data with Completion and Masking (NeurIPS 2025 · similarity 0.88)
- Time Series Generation Under Data Scarcity: A Unified Generative Modeling Approach (NeurIPS 2025 · similarity 0.88)
- TimeDART: A Diffusion Autoregressive Transformer for Self-Supervised Time Series Representation (ICML 2025 · similarity 0.88)
- ABC: Any-Subset Autoregression via Non-Markovian Diffusion Bridges in Continuous Time and Space (NeurIPS 2026 · similarity 0.87)
- Diffusion-TS: Interpretable Diffusion for General Time Series Generation (ICLR 2024 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
