SMART: A Spectral Transfer Approach to Multi-Task Learning
Boxin Zhao, Mladen Kolar, Jinchi Lv · 2026
Read SMART: A Spectral Transfer Approach to Multi-Task Learning by Boxin Zhao, Mladen Kolar, Jinchi Lv (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-2604-20161.txt.
Description
Zhao, Kolar, and Lv propose SMART, a spectral transfer scheme for multi-task regression with small target samples. It assumes spectral likeness, not bounded gaps between source and target models: target singular subspaces sit inside source ones and align sparsely with source bases. SMART builds the target coefficient matrix with structured regularization from source clues, needing only a fitted source model. Though the optimization is nonconvex, an ADMM routine delivers error bounds and a minimax lower bound, with near-minimax rates when the source is noiseless.
Catalog record
| CREATOR | Boxin Zhao, Mladen Kolar, Jinchi Lv |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 186549 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.20161