On the implicit regularization of Langevin dynamics with projected noise

We study Langevin dynamics with noise projected onto the directions orthogonal to an isometric group action. This mathematical model is introduced to shed new light on the effects of symmetry on stochastic gradient descent for over-parametrized models. Our main result identifies a novel form of implicit regularization: when the initial and target density are both invariant under the group action, Langevin dynamics with projected noise is equivalent in law to Langevin dynamics with isotropic diffusion but with an additional drift term proportional to the negative log volume of the group orbit. We prove this result by constructing a coupling of the two processes via a third process on the group itself, and identify the additional drift as the mean curvature of the orbits.

💡 Research Summary

The paper investigates a continuous‑time model of stochastic gradient descent (SGD) in the presence of model over‑parameterization that arises from symmetry. The authors formalize the symmetry by assuming a compact Lie group G ⊂ O(d) acting isometrically on the parameter space ℝᵈ such that the loss function f(x)=E_z