Reproduce “Riemannian Optimization on Unknown Manifolds via Stochastic Proximal Refinement” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Ruixiang Wu, Jiahao Ai, Tinko Sebastian Bartels, Xuanzhuo Liu, Yifei Wu, Yu Mao, Tongxin Li, Costas Courcoubetis
PRISM uses feasible examples to optimize objectives without explicit constraint geometry. It samples candidates using objective value and distance from the current point, then re-anchors on each sample. Under stated assumptions, expected displacement approximates a Riemannian-gradient step, though individual samples need not descend. In BOPTEST, mean weighted objective was 26.969 ± 0.054 versus 32.128 ± 1.102 for budget-matched CEM-MPC, not measured energy savings.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.