← Back to the feed
AI & softwareResearch findings

Nullasar: A Generative Model for Learning Quasar Variability

Original published 2026.11.11 13:59 KSTAdded to Blackdeck 2026.11.15 17:39 KST

What has changed?

Nullasar learns complex statistical properties of quasar light curves without assuming any noise model

Key points

  1. Quasar optical variability is traditionally described by the damped random walk (DRW) model
  2. The DRW model is incomplete and may bias periodic signal detection
  3. Nullasar learns variability directly from time-series data and outperforms parametric fitting in flexible AR(2) models

Connect it to your work

Overcomes limitations of the DRW model by directly learning variability from time-series data, providing a more accurate null hypothesis

What remains unproven

This research note is based on an abstract; check the original for test conditions and application scope.

Originals used for this note

Links open the original publisher in a new tab.

  1. arxivMoving Beyond the DRW Model: Learning Quasar Variability with Generative Models