Sample Complexity of Causal Identification with Temporal Heterogeneity
2026
Read Sample Complexity of Causal Identification with Temporal Heterogeneity (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-2602-06899.txt.
Description
Ameya Rathod and four co-authors ask how much data it takes to pin down a cause-and-effect graph when the world itself shifts over time. From observations alone, many different cause graphs can fit the same data, so the graph is not fixed without extra structure. Past work leaned on either time patterns or on data from many settings, but not on both at once.
The team joins the two kinds of variation and works out shared rules for when the graph can be fixed at all. Time structure, they show, can stand in for missing setting variety. Then they price the job: under heavy-tailed noise, the shape rules stay the same, yet the data needed grows far beyond the bell-curve case. Sharp theory bounds mark the limits of covariance-based recovery for real shifting systems.
Catalog record
| YEAR | 2026 |
|---|---|
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 67 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-06899