Move on Muon : A Hamiltonian probability gradient flow perspective of Muon optimizer

by [unverified] · free · Last verified 2026-06-21T03:07:56.233Z

We develop a gradient flow on the space of probability measures defined on matrix-valued parameters induced by regularized Muon, an analytically smoothed version of the idealized Muon optimizer. The key observation is that the regularized orthogonalization map is the gradient of a smooth Fenchel-...