Abstract:We consider the (sub-Riemannian type) control problem of finding a path going from an initial point $x$ to a target point $y$, by only moving in certain admissible directions. We assume that the corresponding vector fields satisfy the bracket-generating (Hörmander) condition, so that the classical Chow-Rashevskii theorem guarantees the existence of such a path. One natural way to try to solve this problem is via a gradient flow on control space. However, since the corresponding dynamics may have saddle points, any convergence result must rely on suitable (e.g. random) initialisation. We consider the case when this initialisation is irregular, which is conveniently formulated via Lyons' rough path theory. We show that one advantage of this initialisation is that the saddle points are moved to infinity, while minima remain at a finite distance from the starting point. In the step $2$-nilpotent case, we further manage to prove that the gradient flow converges to a solution, if the initial condition is the path of a Brownian motion (or rougher). The proof is based on combining ideas from Malliavin calculus with Łojasiewicz inequalities. A possible motivation for our study comes from the training of deep Residual Neural Nets, in the regime when the number of trainable parameters per layer is smaller than the dimension of the data vector.

A Gradient Flow for the Prescribed Gaussian Curvature Problem on a Closed Riemann Surface with Conical Singularity

Flow by Gauss Curvature to the Aleksandrov and Dual Minkowski Problems

Asymptotic Convergence for a Class of Fully Nonlinear Curvature Flows

Flow by Gauss curvature to the $L_p$-Gaussian Minkowski problem

Surfaces Expanding By The Power Of The Gauss Curvature Flow

A Note on the Mean Curvature Flow in Riemannian Manifolds

Mean Curvature Flow of Arbitrary Codimension in Complex Projective Spaces

Gradient Flow for $\Beta$-Symplectic Critical Surfaces

The Prescribed $Q$-Curvature Flow for Arbitrary Even Dimension in a Critical Case

A Flow Approach to the Prescribed Gaussian Curvature Problem in ℍ𝑛+1

A Variant Prescribed Curvature Flow on Closed Surfaces with Negative Euler Characteristic

Hyperbolic Gradient Flow: Evolution of Graphs in R^n+1

Asymptotic convergence for a class of fully nonlinear inverse curvature flows in a cone

Prescribing Gaussian Curvature on Closed Riemann Surface with Conical Singularity in the Negative Case

Gauss curvature type flow and Alexandrov-Fenchel inequalities in the hyperbolic space

A gradient flow on control space with rough initial condition

A Discrete Ricci Flow on Surfaces with Hyperbolic Background Geometry

Pinching estimates of hypersurfaces by a generalized Gauss curvature flow

Remarks on the 𝐐-curvature Flow

Remarks on the $\mathbf Q$-curvature flow

General curvature flow without singularities