Περίληψη :
We introduce QNPy (https://pypi.org/project/QNPy-Latte/) and QhX (https://pypi.org/project/QhX/), two open-source Python tools for analyzing quasar variability in Vera C. Rubin Observatory-scale time-domain surveys. QNPy combines Self-Organizing Maps and Attentive Latent Neural Processes to learn context-aware latent representations of light curves, enabling the inference of SMBH-related parameters through probabilistic multi-dimensional modeling. QhX, akin to 2D spectroscopy, applies wavelet-based time-frequency transforms to detect complex oscillations and assess their robustness via statistical vetting. Both tools are modular, interoperable, and designed to cooperate with other packages and methodologies. Rather than replacing existing codebases, they are intended to complement ongoing efforts. They support the generation of catalogs of stratified, detected oscillations in quasars — including those potentially arising from close-binary supermassive black hole systems.
Their architecture promotes collaboration within heterogeneous ensemble learning pipelines and aligns with the vision of AI Polymath models, which seek to integrate diverse reasoning methods and representations. The tools have been tested on datasets from the LSST AGN Data Challenge (based on SDSS), as well as ZTF and Gaia.
