Authors: David M. Reiman, John Tamanas, J. Xavier Prochaska, and Dominika Ďurovčíková

In Fully probabilistic quasar continua predictions near Lyman-α with conditional neural spline flows, we introduce Spectre, a conditional rational quadratic neural spline flow which models the distribution over plausible blue-side QSO continua (1190 Å ≤ λ_rest < 1290 Å) given the red-side spectrum (1290 Å ≤ λ_rest < 2900 Å).

In this repository, you'll find:

• Additional reconstruction examples generated from our test set
• Two .npz files each containing 5000 blue-side predictions for each of ULAS J1120+0641 and ULAS J1342+0928
• An inference notebook which loads the optimized Spectre model and predicts blue-side continua given red-side spectra

Required python packages are listed in environment.yml. To create a Conda environment with these dependencies use the following command:

conda env create -f environment.yml

Spectre consumes pre-processed continua from the 14th data release (DR14) of the Sloan Digital Sky Survey (SDSS) quasar catalog captured by the extended Baryon Oscillation Spectroscopic Survey (eBOSS). This repository contains one such datum for use in the inference notebook. The pre-processing code is available here: https://github.com/DominikaDu/QSmooth.

Our code is built upon the Neural Spline Flow codebase whose corresponding paper is available on arXiv at 1906.04032.