{"id":"arxiv/2605.11238","url":"https://arxiv.org/abs/2605.11238","tags":["math.ST","stat.CO"],"score":0,"title":"Efficient Robust Constrained Signal Detection via Kolmogorov Width Approximations","domain":"math.ST","source":"arxiv","authors":["Yikun Li","Matey Neykov"],"reviews":0,"subject":"math.ST","summary":"Robust statistical inference often faces a severe computational-statistical gap when dealing with complex parameter spaces. We investigate minimax signal detection in the Gaussian sequence model under strong $ε$-contamination, where the signal belongs to a general prior constraint $K$. Existing optimal tests require computing the exact Kolmogorov $k$-width of $K$, a computationally intractable task for general non-trivial sets. We bridge this gap by proposing a polynomial-time testing framework that universally applies to balanced, type-2, and exactly 2-convex constraints. By leveraging a…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-05-11","reactions":{}}