{"id":"arxiv/2609.22642","url":"https://arxiv.org/abs/2609.22642","tags":["stat.ML","cs.LG","math.OC","math.ST"],"score":0,"title":"Locally Private Inference for Riemannian Stochastic Optimization","domain":"stat.ML","source":"arxiv","authors":["Xiaotian Chang","Yangdi Jiang","Qirui Hu"],"reviews":0,"subject":"stat.ML","summary":"We develop inference for manifold-valued population minimizers when each observation belongs to a different participant and only locally private messages reach the analyst. The method releases randomized tangent gradients and combines them through Riemannian stochastic approximation and Polyak-Ruppert averaging. Directly inserting a private data surrogate into a nonlinear loss can shift its population target, whereas conditional centring of the released gradient preserves the first-order equation. We introduce symmetric-pair regression (SPR) to estimate the asymptotic variance from the same…","category":"research","comments":0,"disputed":false,"division":"stat","mentions":0,"discussed":0,"published":"2026-09-18","reactions":{}}