qp_flexzboost

This package allows for efficient, lossless storage of Flexcode¹² conditional density estimates and leverages the machinery provided by qp.

The primary module in the package provides the FlexzboostGen class, a subclass of the qp.Pdf_rows_gen class.

An API to retrieve PDF, CDF, and PPF values in addition to supporting simple plotting of PDFs is provided.

While it is possible to use all of the standard scipy.rvs_continuous methods to work with a qp.Ensemble of CDEs stored as FlexzboostGen objects, it is much more efficient to convert the FlexzboostGen representation into a native qp representation, such as qp.interp.

FlexzboostGen is not included as a part of qp by default for the following reasons:

1. It is not possible to convert from a native qp representation into a FlexzboostGen representation because FlexzboostGen stores the output of machine learned model. However, it is possible to convert from FlexzboostGen to any other native qp representation.
2. The use case is very tightly coupled to Flexcode and currently supports one specific use case - efficient storage of qp.Ensemble objects produced as output from rail_flexzboost stages.

For more information and usage examples, please see the documentation and API reference available here: https://qp-flexzboost.readthedocs.io/en/latest/index.html

Attribution

This project was automatically generated using the LINCC Frameworks Python Project Template.

For more information about the project template see the documentation.

Footnotes

1. Rafael Izbicki and Ann B. Lee, “Converting high-dimensional regression to high-dimensional conditional density estimation”, Electron. J. Statist. 11(2): 2800-2831 (2017). DOI: 10.1214/17-EJS1302 ↩

2. Schmidt et al, “Evaluation of probabilistic photometric redshift estimation approaches for The Rubin Observatory Legacy Survey of Space and Time (LSST)“, MNRAS, 449(2): 1587-1606. https://doi.org/10.1093/mnras/staa2799 ↩