TimePre: Bridging Accuracy, Efficiency, and Stability in Probabilistic Time-Series Forecasting

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Jiang, Lingyu, Xu, Lingyu, Li, Peiran, Hou, Dengzhe, Ge, Qianwen, Zhuang, Dingyi, Xing, Shuo, Chen, Wenjing, Gao, Xiangbo, Chen, Ting-Hsuan, Zhan, Xueying, Zhang, Xin, Zhang, Ziming, Tu, Zhengzhong, Zielewski, Michael, Yamada, Kazunori, Lin, Fangzhou
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917430147678208
author Jiang, Lingyu
Xu, Lingyu
Li, Peiran
Hou, Dengzhe
Ge, Qianwen
Zhuang, Dingyi
Xing, Shuo
Chen, Wenjing
Gao, Xiangbo
Chen, Ting-Hsuan
Zhan, Xueying
Zhang, Xin
Zhang, Ziming
Tu, Zhengzhong
Zielewski, Michael
Yamada, Kazunori
Lin, Fangzhou
author_facet Jiang, Lingyu
Xu, Lingyu
Li, Peiran
Hou, Dengzhe
Ge, Qianwen
Zhuang, Dingyi
Xing, Shuo
Chen, Wenjing
Gao, Xiangbo
Chen, Ting-Hsuan
Zhan, Xueying
Zhang, Xin
Zhang, Ziming
Tu, Zhengzhong
Zielewski, Michael
Yamada, Kazunori
Lin, Fangzhou
contents We propose TimePre, a simple framework that unifies the efficiency of Multilayer Perceptron (MLP)-based models with the distributional flexibility of Multiple Choice Learning (MCL) for Probabilistic Time-Series Forecasting (PTSF). Stabilized Instance Normalization (SIN), the core of TimePre, is a normalization layer that explicitly addresses the trade-off among accuracy, efficiency, and stability. SIN stabilizes the hybrid architecture by correcting channel-wise statistical shifts, thereby resolving the catastrophic hypothesis collapse. Extensive experiments on six benchmark datasets demonstrate that TimePre achieves state-of-the-art (SOTA) accuracy on key probabilistic metrics. Critically, TimePre achieves inference speeds that are orders of magnitude faster than sampling-based models, and is more stable than prior MCL approaches.
format Preprint
id arxiv_https___arxiv_org_abs_2511_18539
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle TimePre: Bridging Accuracy, Efficiency, and Stability in Probabilistic Time-Series Forecasting
Jiang, Lingyu
Xu, Lingyu
Li, Peiran
Hou, Dengzhe
Ge, Qianwen
Zhuang, Dingyi
Xing, Shuo
Chen, Wenjing
Gao, Xiangbo
Chen, Ting-Hsuan
Zhan, Xueying
Zhang, Xin
Zhang, Ziming
Tu, Zhengzhong
Zielewski, Michael
Yamada, Kazunori
Lin, Fangzhou
Machine Learning
Computer Vision and Pattern Recognition
We propose TimePre, a simple framework that unifies the efficiency of Multilayer Perceptron (MLP)-based models with the distributional flexibility of Multiple Choice Learning (MCL) for Probabilistic Time-Series Forecasting (PTSF). Stabilized Instance Normalization (SIN), the core of TimePre, is a normalization layer that explicitly addresses the trade-off among accuracy, efficiency, and stability. SIN stabilizes the hybrid architecture by correcting channel-wise statistical shifts, thereby resolving the catastrophic hypothesis collapse. Extensive experiments on six benchmark datasets demonstrate that TimePre achieves state-of-the-art (SOTA) accuracy on key probabilistic metrics. Critically, TimePre achieves inference speeds that are orders of magnitude faster than sampling-based models, and is more stable than prior MCL approaches.
title TimePre: Bridging Accuracy, Efficiency, and Stability in Probabilistic Time-Series Forecasting
topic Machine Learning
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2511.18539