research
At Realistic Noise, Simple Shapes Catch the Neural Network
On 10,000 simulated thermal Sunyaev-Zel’dovich maps, interpretable summaries matched CNN constraints once observatory-like noise was added.
Summary
On 10,000 simulated thermal Sunyaev-Zel’dovich maps, interpretable summaries matched CNN constraints once observatory-like noise was added.
The first field-level analysis in the study found tighter noiseless constraints from a convolutional network, but Minkowski functionals combined with the power spectrum became comparable under Simons Observatory-like noise. The comparison uses simulations and motivates real-data work rather than replacing it.
Why it matters
On 10,000 simulated thermal Sunyaev-Zel’dovich maps, interpretable summaries matched CNN constraints once observatory-like noise was added.
Limits and context
No additional limitation was separately recorded.
Key claims
On 10,000 simulated thermal Sunyaev-Zel’dovich maps, interpretable summaries matched CNN constraints once observatory-like noise was added.
Evidence: source-2026-10-30-020
Sources
- arXiv preprint 2610.40617, original v1arXiv · primary research
Corrections
No corrections have been recorded for this story.