mli0118/riemannian-optimization

This repository contains research experiments for deterministic and stochastic Riemannian bilevel optimization, rather than a general-purpose application. The BCI experiments implement Grassmann, Stiefel, and synthetic hyperbolic variants; the stochastic experiments evaluate three manifolds on UCI Superconductivity data, comparing SR-HJFBiO with hypergradient baselines. The main workflow is to load or generate data, construct manifold-specific features, run an optimizer, save trajectories, and plot or compare results. The README is empty, and sampled excerpts are partial, so links between unsampled experiment internals are omitted where their wiring is unclear.