Coefficient-to-Basis Network: A Fine-Tunable Operator Learning Framework for Inverse Problems with Adaptive Discretizations and Theoretical Guarantees
Zecheng Zhang, Hao Liu, Wenjing Liao, Guang Lin · 2025
Read Coefficient-to-Basis Network: A Fine-Tunable Operator Learning Framework for Inverse Problems with Adaptive Discretizations and Theoretical Guarantees by Zecheng Zhang, Hao Liu, Wenjing Liao, Guang Lin (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-2503-08642.txt.
Description
Zecheng Zhang and co-authors present a network that solves inverse problems and adapts to new grids by fine-tuning. Inverse problems infer hidden causes from observed effects. Most learning methods need full retraining when the grid changes.
Their Coefficient-to-Basis Network starts from a pretrained model. Light fine-tuning then fits it to new discretizations at low cost. Accuracy stays high without training from scratch each time. Theory gives error bounds for approximation and generalization. The analysis shows the network finds low-dimensional structure on its own. No explicit coding of that structure is required. Tests on several inverse problems beat prior methods. The work cuts compute cost for changing sensor layouts.
Catalog record
| CREATOR | Zecheng Zhang, Hao Liu, Wenjing Liao, Guang Lin |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 62 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2503-08642