Routing Channel-Patch Dependencies in Time Series Forecasting with Graph Spectral Decomposition
Dongyuan Li, Shun Zheng, Chang Xu, Jiang Bian, Renhe Jiang · 2026
Read Routing Channel-Patch Dependencies in Time Series Forecasting with Graph Spectral Decomposition by Dongyuan Li, Shun Zheng, Chang Xu, Jiang Bian, Renhe Jiang (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-2603-13702.txt.
Description
Dongyuan Li and co-authors propose xCPD, a plugin for routing channel-patch dependencies in time series forecasting. Channel-Independent modeling treats each channel alone, which can miss inter-channel interactions and generalize poorly. Channel-Dependent modeling pools all channels, which can add irrelevant data and oversmooth. The plugin projects multivariate signals into the frequency domain with a shared graph Fourier basis. Patches are grouped into low-, mid-, and high-frequency bands by spectral energy. A channel-adaptive routing mechanism then sets inter-channel interaction per patch.
Catalog record
| CREATOR | Dongyuan Li, Shun Zheng, Chang Xu, Jiang Bian, Renhe Jiang |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 93544 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2603.13702