EzTaoX

EzTaoX is a flexible framework for multi-wavelength and multi-survey AGN light-curve modeling using Gaussian Processes (GPs). Built on top of tinygp---a scalable GP library in JAX---EzTaoX is fast, modular, and integrates seamlessly with the JAX ecosystem for statistical inference and modern machine learning.

EzTaoX is under active development, breaking API changes are expected.

Installation

pip install eztaox

Dependencies

Supports Python 3.10, 3.11, 3.12

"jax (<=0.4.31)",
"jaxlib (<=0.4.31)",
"tinygp (>=0.3.0,<0.4.0)",
"jaxopt (>=0.8.3,<0.9.0)",
"optax (>=0.2.4,<0.3.0)",
"numpyro (>=0.17.0,<0.20.0)",

Documentation

Please see our readthedocs

Citation

If you find EzTaoX useful for your research, please consider citing the following paper arXiv:2511.21479:

@ARTICLE{Yu2025,
       author = {{Yu}, Weixiang and {Ruan}, John J. and {Burke}, Colin J. and {Assef}, Roberto J. and {Ananna}, Tonima T. and {Bauer}, Franz E. and {De Cicco}, Demetra and {Horne}, Keith and {Hern{\'a}ndez-Garc{\'\i}a}, Lorena and {Ili{\'c}}, Dragana and {Kova{\v{c}}evi{\'c}}, Andjelka B. and {Marculewicz}, Marcin and {Panda}, Swayamtrupta and {Ricci}, Claudio and {Richards}, Gordon T. and {Riffel}, Rogemar A. and {Schneider}, Donald P. and {S{\'a}nchez-S{\'a}ez}, Paula and {Satheesh Sheeba}, Sarath and {Tombesi}, Francesco and {Temple}, Matthew J. and {Vogeley}, Michael S. and {Yoon}, Ilsang and {Zou}, Fan},
        title = "{Scalable and Robust Multiband Modeling of AGN Light Curves in Rubin-LSST}",
      journal = {arXiv e-prints},
     keywords = {Astrophysics of Galaxies, Instrumentation and Methods for Astrophysics},
         year = 2025,
        month = nov,
          eid = {arXiv:2511.21479},
        pages = {arXiv:2511.21479},
          doi = {10.48550/arXiv.2511.21479},
archivePrefix = {arXiv},
       eprint = {2511.21479},
 primaryClass = {astro-ph.GA},
       adsurl = {https://ui.adsabs.harvard.edu/abs/2025arXiv251121479Y},
      adsnote = {Provided by the SAO/NASA Astrophysics Data System}
}

Acknowledgment

EzTaoX is built on top of (and inspired by) tinygp---a general purpose GP modeling framework written in JAX. For more general GP modeling tasks, experienced users can directly explore tinygp.