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

2025-11-23

AI Summary

Existing probabilistic time series forecasting (PTSF) methods struggle to simultaneously achieve accuracy, efficiency, and training stability: diffusion-based generative models incur prohibitive inference costs, while efficient non-sampling frameworks—such as multi-choice learning (MCL)—suffer from catastrophic hypothesis collapse and training instability, especially when paired with MLP backbones. This paper proposes a novel PTSF framework centered on Stable Instance Normalization (SIN), a channel-wise statistical calibration layer that effectively mitigates hypothesis collapse in MCL. SIN enables the first stable, efficient, multi-hypothesis probabilistic forecasting with MLP architectures. The method eliminates iterative sampling and supports end-to-end optimization. Evaluated on six benchmark datasets, it establishes new state-of-the-art probabilistic forecasting performance, achieves inference speedups of several orders of magnitude over diffusion models, and demonstrates robust training stability and scalability.

Technology Category

• Machine Learning: Probabilistic Circuits and Graphical Models
• Reasoning under Uncertainty: Relational Probabilistic Models
• Search and Optimization: Sampling/Simulation-based Search

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Large language models for search

Abstract

Probabilistic Time-Series Forecasting (PTSF) is critical for uncertainty-aware decision making, but existing generative models, such as diffusion-based approaches, are computationally prohibitive due to expensive iterative sampling. Non-sampling frameworks like Multiple Choice Learning (MCL) offer an efficient alternative, but suffer from severe training instability and hypothesis collapse, which has historically hindered their performance. This problem is dramatically exacerbated when attempting to combine them with modern, efficient MLP-based backbones. To resolve this fundamental incompatibility, we propose TimePre, a novel framework that successfully unifies the efficiency of MLP-based models with the distributional flexibility of the MCL paradigm. The core of our solution is Stabilized Instance Normalization (SIN), a novel normalization layer that explicitly remedies this incompatibility. SIN stabilizes the hybrid architecture by correcting channel-wise statistical shifts, definitively resolving the catastrophic hypothesis collapse. Extensive experiments on six benchmark datasets demonstrate that TimePre achieves new state-of-the-art accuracy on key probabilistic metrics. Critically, TimePre achieves inference speeds orders of magnitude faster than sampling-based models and, unlike prior MCL work, demonstrates stable performance scaling. It thus bridges the long-standing gap between accuracy, efficiency, and stability in probabilistic forecasting.

Problem

• Resolving training instability in non-sampling probabilistic forecasting models
• Combining MLP efficiency with distributional flexibility in time-series
• Bridging the gap between accuracy, efficiency, and forecasting stability

Innovation

• Stabilized Instance Normalization stabilizes hybrid architecture
• Unifies MLP efficiency with MCL distributional flexibility
• Achieves state-of-the-art accuracy with fast inference speeds