Linear Algebra Problems Solved by Using Damped Dynamical Systems on the Stiefel Manifold
This work provides a new optimization approach for problems on the Stiefel manifold, relevant for applications in machine learning and signal processing, but the improvements over existing methods are incremental.
The authors propose a novel method for solving minimization problems on the Stiefel manifold by combining damped dynamical systems, with constraints satisfied in the limit. Numerical experiments show competitive performance compared to a state-of-the-art conjugate gradient method.
We develop a new method for solving minimization problems on the Stiefel Manifold using damped dynamical systems. The constraints are satisfied in the limit by an additional damped dynamical system. The method is illustrated by numerical experiments and compared to a state-of-the-art conjugate gradient method.