New Method Offers Confidence Regions for Riemannian Stochastic Gradient Estimates
A new paper proposes self-normalized inference for Riemannian stochastic gradient paths, enabling confidence regions for constrained parameters without estimating the Hessian and gradient-noise covariance.
Read the full story at arXiv stat.ML ↗
Timeline · 1 report
- 2026-11-03 05:00 · arXiv stat.ML
Self-Normalized Inference from Riemannian Stochastic Gradient Paths