Domain Fusion Controllable Generalization for Cross-Domain Time Series Forecasting from Multi-Domain Integrated Distribution

Xiangkai Ma, Xiaobin Hong, Mingkai Lin, Han Zhang, Wenzhong Li, Sanglu Lu · 2024

Read Domain Fusion Controllable Generalization for Cross-Domain Time Series Forecasting from Multi-Domain Integrated Distribution by Xiangkai Ma, Xiaobin Hong, Mingkai Lin, Han Zhang, Wenzhong Li, Sanglu Lu (2024) 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-2412-03068.txt.

Description

Ma and co-authors present in 2024 TimeControl, a diffusion model that generalizes time series forecasting across domains. Deep forecasters shine in one domain but shift badly to others, and prompt tricks using statistics fail under large shifts. Their domain fusion design folds many domains into one generative process.

A condition network reads multi scale patterns from the input series and guides denoising toward the forecast. Pretraining learns a shared representation, then adapter tuning adapts it to a target domain. A hybrid architecture aligns inputs across domains. Forecasts are sampled directly for the target, rather than rolled out step by step.

Catalog record

CREATORXiangkai Ma, Xiaobin Hong, Mingkai Lin, Han Zhang, Wenzhong Li, Sanglu Lu
YEAR2024
COLLECTIONPAPERS
FORMATTXT
FILE SIZE138 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2412-03068

Browse the full library · About the Tower of Babel