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

Transformers have been actively studied for time-series forecasting in recentyears. While often showing promising results in various scenarios, traditionalTransformers are not designed to fully exploit the characteristics oftime-series data and thus suffer some fundamental limitations, e.g., theygenerally lack of decomposition capability and interpretability, and areneither effective nor efficient for long-term forecasting. In this paper, wepropose ETSFormer, a novel time-series Transformer architecture, which exploitsthe principle of exponential smoothing in improving Transformers fortime-series forecasting. In particular, inspired by the classical exponentialsmoothing methods in time-series forecasting, we propose the novel exponentialsmoothing attention (ESA) and frequency attention (FA) to replace theself-attention mechanism in vanilla Transformers, thus improving both accuracyand efficiency. Based on these, we redesign the Transformer architecture withmodular decomposition blocks such that it can learn to decompose thetime-series data into interpretable time-series components such as level,growth and seasonality. Extensive experiments on various time-series benchmarksvalidate the efficacy and advantages of the proposed method. Code is availableat https://github.com/salesforce/ETSformer.

ETSformer: Exponential Smoothing Transformers for Time-series Forecasting