Non-stationary Transformers: Exploring the Stationarity in Time Series Forecasting

By Yong Liu, Haixu Wu, Jianmin Wang, Mingsheng Long. Neural Information Processing Systems, 2022

940 citations recorded by the source corpus.

Transformers have shown great power in time series forecasting due to their global-range modeling ability. However, their performance can degenerate terribly on non-stationary real-world data in which the joint distribution changes over time. Previous studies primarily adopt stationarization to attenuate the non-stationarity of original series for better predictability. But the stationarized series deprived of inherent non-stationarity can be less instructive for real-world bursty events forecasting. This problem, termed over-stationarization in this paper, leads Transformers to generate indistinguishable temporal attentions for different series and impedes the predictive capability of deep models. To tackle the dilemma between series predictability and model capability, we propose Non-stationary Transformers as a generic framework with two interdependent modules: Series Stationarization and De-stationary Attention. Concretely, Series Stationarization unifies the statistics of each input and converts the output with restored statistics for better predictability. To address the over-stationarization problem, De-stationary Attention is devised to recover the intrinsic non-stationary information into temporal dependencies by approximating distinguishable attentions learned from raw series. Our Non-stationary Transformers framework consistently boosts mainstream Transformers by a large margin, which reduces MSE by 49.43% on Transformer, 47.34% on Informer, and 46.89% on Reformer, making them the state-of-the-art in time series forecasting. Code is available at this repository: https://github.com/thuml/Nonstationary_Transformers.

Subfields: Applied Mathematics, Management Science and Operations Research, Signal Processing

View Paper: https://www.semanticscholar.org/paper/8064d3873c646dc9ff949d72c54c634a906fc092

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Management Science and Operations Research

1. Tomas Pfister
2. Nate Yoder
3. Chun-Liang Li
4. Si Chen
5. Lintao Ma
6. Shiyu Wang
7. Sercan Ö. Arik

In Signal Processing

1. Tomas Pfister
2. Nate Yoder
3. Chun-Liang Li
4. Xiaoming Shi
5. Si Chen
6. Lintao Ma
7. Shiyu Wang
8. Sercan Ö. Arik

Citations in the corpus

• TSMixer: An all-MLP Architecture for Time Series Forecasting (2023, methodology): “…have been proposed due to their capacity to capture both complex temporal patterns and cross-variate dependencies (Gamboa, 2017; Li et al., 2019; Wen et al., 2017; Zhou et al., 2021; Wu et al., 2021; Lim & Zohren, 2021; Liu et al., 2022a; Zhou et al., 2022b; Liu et al., 2022b; Zhou et al., 2022a) .”
• Time-LLM: Time Series Forecasting by Reprogramming Large Language Models (2023, methodology): “We mainly follow the experimental configurations in (Wu et al., 2022) across all baselines within a unified evaluation pipeline in https://github.com/thuml/Time-Series-Library for fair comparisons.”
• TRACE: Grounding Time Series in Context for Multimodal Embedding and Retrieval (2025, unknown): “Transformer-based models leverage self-attention to capture long-range dependencies and flexible temporal dynamics [19, 20, 21, 22, 23, 24, 25, 26, 27, 28].”