Monthly Notices of the Royal Astronomical Society · 23 October 2026 · From issue 56
Uncertainty-Aware Deep Learning for the Lyα Forest: CNN-Based Absorber Detection and Characterization
A fast computer program can read gas clouds in quasar light almost as well as slow older methods, in simulations.
Paryag Sharma, Vikram Khaire et al.
Light from distant quasars passes through gas clouds in deep space, leaving patterns called absorption lines. Scientists built a computer program that reads these patterns fast, using a method inspired by how brains process images. It trained on fake star light made by a computer simulation. It matched an older, slower method about 80% of the time on that fake data. On real telescope data it only matched about 50% of the time, because real light looks different from the practice data.
How often the new fast method matched the old, slow method for finding gas clouds in quasar light.
Simulation-trained model only, real telescope accuracy dropped a lot.
Integrity screen: passed (3 checks) Checked 24 October 2026. Retraction record: none. DOI resolves at doi.org. Metadata record found (Monthly Notices of the Royal Astronomical Society). Read the source
This screen checks process signals. It does not tell you the findings are true.