One Rank at a Time: Cascading Error Dynamics in Sequential Learning
Mahtab Alizadeh Vandchali, Fangshuo (Jasper) Liao, Anastasios Kyrillidis · 2025
Read One Rank at a Time: Cascading Error Dynamics in Sequential Learning by Mahtab Alizadeh Vandchali, Fangshuo (Jasper) Liao, Anastasios Kyrillidis (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-2505-22602.txt.
Description
Mahtab Alizadeh Vandchali, Fangshuo Liao, and Anastasios Kyrillidis wrote this 2025 paper on how errors grow in sequential learning. In that approach, a hard task is split into simpler parts that are learned one after another. The authors study the process through low-rank linear regression, where each step learns a rank-1 piece and rests on the steps before it.
The paper builds a framework that breaks the full process into a chain of small estimation problems. It then sets bounds on how errors from tight budgets and limited precision spread through the chain. The results show that these errors build up in ways that can be predicted. That gives designers a clearer guide for keeping such systems stable and accurate.
Catalog record
| CREATOR | Mahtab Alizadeh Vandchali, Fangshuo (Jasper) Liao, Anastasios Kyrillidis |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 107 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2505-22602