Dynamic Perturbed Adaptive Method for Infinite Task-Conflicting Time Series
Jiang YOU Department of Computer ScienceUniversité Gustave Eiffel & ESIEE-Paris [email protected], Xiaozhen WANG CEREMADEUniversité Paris-Dauphine PSL [email protected] Arben CELADepartment of Computer ScienceUniversité Gustave Eiffel & ESIEE-Paris [email protected] · 2025
Read Dynamic Perturbed Adaptive Method for Infinite Task-Conflicting Time Series by Jiang YOU Department of Computer ScienceUniversité Gustave Eiffel & ESIEE-Paris [email protected], Xiaozhen WANG CEREMADEUniversité Paris-Dauphine PSL [email protected] Arben CELADepartment of Computer ScienceUniversité Gustave Eiffel & ESIEE-Paris [email protected] (2025) 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-2505-11902.txt.
Description
Jiang You, Xiaozhen Wang, and Arben Cela proposed a learning method in twenty twenty-five for time series whose task keeps changing. In their setup, the same input can demand different outputs under shifting goals, which breaks static models. They built a synthetic data set full of conflicting subtasks to stress-test adaptation under frequent shifts.
Their method splits the model into a trunk and branches. The trunk changes slowly and holds long-term structure. Each branch is reset and retrained for the current task, letting the system adapt at test time without task labels. They prove this design is strictly more expressive than static models and a popular tuning method, and that branch adaptation converges fast under a known condition. Experiments beat strong baselines in clashing-task settings.
Catalog record
| CREATOR | Jiang YOU Department of Computer ScienceUniversité Gustave Eiffel & ESIEE-Paris [email protected], Xiaozhen WANG CEREMADEUniversité Paris-Dauphine PSL [email protected] Arben CELADepartment of Computer ScienceUniversité Gustave Eiffel & ESIEE-Paris [email protected] |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 39 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2505-11902