.png){kind=icon}
.png){kind=icon}
Return
Diffinformer: Diffusion informer model for long sequence time-series forecasting
.png){kind=icon}
PRE
AI
DOI：10.1016/j.eswa.2025.129944.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
• Introduction of Diffinformer: We propose Diffinformer, the first model to combine ProbSparse attention mechanisms with diffusion processes for time series forecasting. • Conditional Diffusion Application: The application of conditional diffusion in time series forecasting enhances the signal-to-noise ratio (SNR), providing theoretical and practical improvements in prediction accuracy. • Efficient and Accurate Predictions: Diffinformer generates full sequence forecasts in a single step, significantly reducing computational complexity while outperforming baseline models across multiple large-scale datasets.
Journal
IF:
7.5
Papers:
2.9W
Citations:
10.2W