DistDF: Time-Series Forecasting Needs Joint-Distribution Wasserstein Alignment
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2025
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866914464376291328 |
|---|---|
| author | Wang, Hao Pan, Licheng Lu, Yuan Chu, Zhixuan Li, Xiaoxi He, Shuting Chen, Zhichao Li, Haoxuan Wen, Qingsong Lin, Zhouchen |
| author_facet | Wang, Hao Pan, Licheng Lu, Yuan Chu, Zhixuan Li, Xiaoxi He, Shuting Chen, Zhichao Li, Haoxuan Wen, Qingsong Lin, Zhouchen |
| contents | Training time-series forecasting models requires aligning the conditional distribution of model forecasts with that of the label sequence. The standard direct forecast (DF) approach resorts to minimizing the conditional negative log-likelihood, typically estimated by the mean squared error. However, this estimation proves biased when the label sequence exhibits autocorrelation. In this paper, we propose DistDF, which achieves alignment by minimizing a distributional discrepancy between the conditional distributions of forecast and label sequences. Since such conditional discrepancies are difficult to estimate from finite time-series observations, we introduce a joint-distribution Wasserstein discrepancy for time-series forecasting, which provably upper bounds the conditional discrepancy of interest. The proposed discrepancy is tractable, differentiable, and readily compatible with gradient-based optimization. Extensive experiments show that DistDF improves diverse forecasting models and achieves leading performance. Code is available at https://anonymous.4open.science/r/DistDF-F66B. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2510_24574 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | DistDF: Time-Series Forecasting Needs Joint-Distribution Wasserstein Alignment Wang, Hao Pan, Licheng Lu, Yuan Chu, Zhixuan Li, Xiaoxi He, Shuting Chen, Zhichao Li, Haoxuan Wen, Qingsong Lin, Zhouchen Machine Learning Artificial Intelligence Training time-series forecasting models requires aligning the conditional distribution of model forecasts with that of the label sequence. The standard direct forecast (DF) approach resorts to minimizing the conditional negative log-likelihood, typically estimated by the mean squared error. However, this estimation proves biased when the label sequence exhibits autocorrelation. In this paper, we propose DistDF, which achieves alignment by minimizing a distributional discrepancy between the conditional distributions of forecast and label sequences. Since such conditional discrepancies are difficult to estimate from finite time-series observations, we introduce a joint-distribution Wasserstein discrepancy for time-series forecasting, which provably upper bounds the conditional discrepancy of interest. The proposed discrepancy is tractable, differentiable, and readily compatible with gradient-based optimization. Extensive experiments show that DistDF improves diverse forecasting models and achieves leading performance. Code is available at https://anonymous.4open.science/r/DistDF-F66B. |
| title | DistDF: Time-Series Forecasting Needs Joint-Distribution Wasserstein Alignment |
| topic | Machine Learning Artificial Intelligence |
| url | https://arxiv.org/abs/2510.24574 |