{"id":"arxiv/2611.11872","url":"https://arxiv.org/abs/2611.11872","tags":["stat.ME","eess.SP","math.ST"],"score":0,"title":"Distribution-Free Symmetry Detection in Multi-Reference Alignment at Low SNR","domain":"stat.ME","source":"arxiv","authors":["Nir Sharon","Sheng Xu"],"reviews":0,"subject":"stat.ME","summary":"Symmetry can reveal the structure of a signal, but detecting it becomes difficult when observations are noisy and randomly transformed. We study how to infer symmetry directly from multi-reference alignment data with additive isotropic Gaussian noise of known variance, without aligning the observations or knowing how the transformations were sampled. For cyclic symmetry, we identify a simple statistical signature that yields tests with exact finite-sample error guarantees. We demonstrate their performance even at very low signal-to-noise ratios. Dihedral symmetry presents a deeper challenge:…","category":"research","comments":0,"disputed":false,"division":"stat","mentions":0,"discussed":0,"published":"2026-11-09","reactions":{}}