A space-decoupling framework for optimization on bounded-rank matrices with orthogonally invariant constraints

TL;DR

This paper introduces a space-decoupling framework that simplifies optimization on bounded-rank matrices with orthogonally invariant constraints, enabling efficient Riemannian algorithms and demonstrating superior performance in various real-world applications.

Contribution

The authors propose a novel space-decoupling approach that transforms complex coupled constraints into a smooth manifold, facilitating optimization and extending applicability to multiple domains.

Findings

Decoupling constraints simplifies the optimization geometry.

The reformulated problem is equivalent to the original.

Numerical experiments show improved performance across applications.

Abstract

Imposing additional constraints on low-rank optimization has garnered growing interest. However, the geometry of coupled constraints hampers the well-developed low-rank structure and makes the problem intricate. To this end, we propose a space-decoupling framework for optimization on bounded-rank matrices with orthogonally invariant constraints. The "space-decoupling" is reflected in several ways. We show that the tangent cone of coupled constraints is the intersection of tangent cones of each constraint. Moreover, we decouple the intertwined bounded-rank and orthogonally invariant constraints into two spaces, leading to optimization on a smooth manifold. Implementing Riemannian algorithms on this manifold is painless as long as the geometry of additional constraints is known. In addition, we unveil the equivalence between the reformulated problem and the original problem. Numerical experiments on real-world applications -- spherical data fitting, graph similarity measuring, low-rank SDP, model reduction of Markov processes, reinforcement learning, and deep learning -- validate the superiority of the proposed framework.