Physically Constrained Causal Noise Models For High-contrast Imaging Of Exoplanets


Finding exoplanets in HCI data requires a multi-stage post-processing pipeline. The most crucial step, however, is the estimation of the stellar PSF.

The detection of exoplanets in high-contrast imaging (HCI) data hinges on post-processing methods to remove spurious light from the host star.

So far, existing methods for this task hardly utilize any of the available domain knowledge about the problem explicitly. We propose a new approach to HCI post-processing based on a modified half-sibling regression scheme, and show how we use this framework to combine machine learning with existing scientific domain knowledge. On three real data sets, we demonstrate that the resulting system performs up to a factor of 4 times better than one of the currently leading algorithms. This has the potential to allow significant discoveries of exoplanets both in new and archival data.

Timothy D. Gebhard, Markus J. Bonse, Sascha P. Quanz, Bernhard Schölkopf
Comments: Submitted to the "Machine Learning and the Physical Sciences" workshop at NeurIPS 2020. 8 pages, 6 figures
Subjects: Instrumentation and Methods for Astrophysics (astro-ph.IM); Machine Learning (cs.LG)
Cite as: arXiv:2010.05591 [astro-ph.IM] (or arXiv:2010.05591v1 [astro-ph.IM] for this version)
Submission history
From: Timothy D. Gebhard
[v1] Mon, 12 Oct 2020 10:35:03 UTC (603 KB)
https://arxiv.org/abs/2010.05591
Astrobiology

Please follow Astrobiology on Twitter.


  • submit to reddit