Interpreting deep learning-based stellar mass estimation via causal analysis and mutual information decomposition
Wei Zhang (张伟), Qiufan Lin (林秋帆), Yuan-Sen Ting (丁源森), Shupei Chen (陈树沛), Hengxin Ruan (阮恒心), Song Li (李松), Yifan Wang (王一凡) · 2025
Read Interpreting deep learning-based stellar mass estimation via causal analysis and mutual information decomposition by Wei Zhang (张伟), Qiufan Lin (林秋帆), Yuan-Sen Ting (丁源森), Shupei Chen (陈树沛), Hengxin Ruan (阮恒心), Song Li (李松), Yifan Wang (王一凡) (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-2509-23901.txt.
Description
A deep network can guess a galaxy's stellar mass from its pictures, but which visual clues actually cause the estimate? Wei Zhang and 6 co-authors interpret such a model with causal analysis and mutual information decomposition. Causal paths separate real influence from mere association among variables. Information decomposition then splits the inputs' contributions into redundant, unique, and synergistic parts.
Data come from the Sloan Digital Sky Survey and the WISE infrared survey. The results give physical readings of how morphology and other image content beyond total brightness feed the mass estimate. The authors argue the pairing of deep learning with interpretability tools can sharpen data-driven astrophysics.
Catalog record
| CREATOR | Wei Zhang (张伟), Qiufan Lin (林秋帆), Yuan-Sen Ting (丁源森), Shupei Chen (陈树沛), Hengxin Ruan (阮恒心), Song Li (李松), Yifan Wang (王一凡) |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 137 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2509-23901