Conformal changepoint localization
Rohan Hore, Aaditya Ramdas · 2026
Read Conformal changepoint localization by Rohan Hore, Aaditya Ramdas (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-2602-06267.txt.
Description
Finding the exact point where a data stream changes character, and saying how sure you are, is the changepoint problem. Rohan Hore and Aaditya Ramdas study it with almost no assumptions: data before and after one change are drawn from unknown distributions, with only exchangeability assumed. The goal is a confidence set for the change index with a finite-sample guarantee.
Their method, CONCH, builds such sets from conformal inference arguments. A new lemma in the Neyman-Pearson style yields score functions that keep the sets small, and the sets shrink toward the true point under weak conditions. They also prove any assumption-free method must be a version of CONCH. Tests included image and text data.
Catalog record
| CREATOR | Rohan Hore, Aaditya Ramdas |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 125 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-06267