XLinear: Frequency-Enhanced MLP with CrossFilter for Robust Long-Range Forecasting

1 st Xiang Ao · 2026

Read XLinear: Frequency-Enhanced MLP with CrossFilter for Robust Long-Range Forecasting by 1 st Xiang Ao (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2603-15645.txt.

Description

Ao proposes XLinear, a small neural forecaster built for long-range time series. Plain MLP forecasters resist noise well but miss long-range links. XLinear first splits a series into trend and seasonal parts. For the trend part, Enhanced Frequency Attention works in the frequency domain. For the seasonal part, a CrossFilter block guards noise robustness. The author reports state-of-the-art scores on test datasets. The model stays light while beating other MLP forecasters on long-range skill.

Catalog record

CREATOR1 st Xiang Ao
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE33547
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2603.15645

Browse the full library · About the Tower of Babel