English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Time series forecasting is a crucial task in various domains. Caused byfactors such as trends, seasonality, or irregular fluctuations, time seriesoften exhibits non-stationary. It obstructs stable feature propagation throughdeep layers, disrupts feature distributions, and complicates learning datadistribution changes. As a result, many existing models struggle to capture theunderlying patterns, leading to degraded forecasting performance. In thisstudy, we tackle the challenge of non-stationarity in time series forecastingwith our proposed framework called U-Mixer. By combining Unet and Mixer,U-Mixer effectively captures local temporal dependencies between differentpatches and channels separately to avoid the influence of distributionvariations among channels, and merge low- and high-levels features to obtaincomprehensive data representations. The key contribution is a novelstationarity correction method, explicitly restoring data distribution byconstraining the difference in stationarity between the data before and aftermodel processing to restore the non-stationarity information, while ensuringthe temporal dependencies are preserved. Through extensive experiments onvarious real-world time series datasets, U-Mixer demonstrates its effectivenessand robustness, and achieves 14.5\% and 7.7\% improvements overstate-of-the-art (SOTA) methods.

U-Mixer: An Unet-Mixer Architecture with Stationarity Correction for Time Series Forecasting