{"id":"arxiv/2609.06637","url":"https://arxiv.org/abs/2609.06637","tags":["math.ST","eess.SP"],"score":0,"title":"Global minimax risk and acquisition laws for heterogeneous information fusion","domain":"math.ST","source":"arxiv","authors":["Armon Rasooli","Mohammad Sadegh Narimani"],"reviews":0,"subject":"math.ST","summary":"We prove an all-allocation global target-risk theorem for independent Gaussian sources that share a scalar nuisance and an unknown contact coordinate. For a fixed immersed nuisance curve with finitely many multiple fibres and pairwise nonparallel branch tangents, squared target risk is comparable to a primary estimation floor plus a target-gap-weighted Gaussian discrimination profile. Independent localization, finite branch selection and target-class refitting give an estimator attaining this comparison across every nonnegative integer allocation. A full-box specialization has…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-09-06","reactions":{}}