Scalable Gaussian Processes for Integrated and Overlapping Measurements Via Augmented State Space Models tinygp ( 19 ) , JAX ( 6 ) , matplotlib ( 25 ) ,
Ryan A. Rubenzahl, Soichiro Hattori, Simo Särkkä, Will M. Farr, Jacob K. Luhn, Megan Bedell · 2026
Read Scalable Gaussian Processes for Integrated and Overlapping Measurements Via Augmented State Space Models tinygp ( 19 ) , JAX ( 6 ) , matplotlib ( 25 ) , by Ryan A. Rubenzahl, Soichiro Hattori, Simo Särkkä, Will M. Farr, Jacob K. Luhn, Megan Bedell (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-2601-02527.txt.
Description
Ryan A. Rubenzahl and six co authors make Gaussian process modeling fast for astronomical measurements that are integrated over an exposure, rather than taken at an instant. Integration blurs fast variability, and once exposure times vary or overlap, the standard covariance matrix loses the structure that makes it cheap. Forcing memory and time costs that grow far too fast with data size.
Linear Gaussian state space models are known to be equivalent to Gaussian processes and can be solved in linear time with Kalman filtering and smoothing. The authors extend that equivalence to integrated measurements by adding an integral state that resets when an exposure starts and is read when it ends. The augmented model gives exactly the same posterior as the full integrated process, in linear time on a CPU and faster still in parallel on a GPU.
Catalog record
| CREATOR | Ryan A. Rubenzahl, Soichiro Hattori, Simo Särkkä, Will M. Farr, Jacob K. Luhn, Megan Bedell |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 76 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2601-02527