Abstract:There are two widely used models for the Grassmannian $\operatorname{Gr}(k,n)$, as the set of equivalence classes of orthogonal matrices $\operatorname{O}(n)/(\operatorname{O}(k) \times \operatorname{O}(n-k))$, and as the set of trace-$k$ projection matrices $\{P \in \mathbb{R}^{n \times n} : P^{\mathsf{T}} = P = P^2,\; \operatorname{tr}(P) = k\}$. The former, standard in manifold optimization, has the advantage of giving numerically stable algorithms but the disadvantage of having to work with equivalence classes of matrices. The latter, widely used in coding theory and probability, has the advantage of using actual matrices (as opposed to equivalence classes) but working with projection matrices is numerically unstable. We present an alternative that has both advantages and suffers from neither of the disadvantages; by representing $k$-dimensional subspaces as symmetric orthogonal matrices of trace $2k-n$, we obtain \[ \operatorname{Gr}(k,n) \cong \{Q \in \operatorname{O}(n) : Q^{\mathsf{T}} = Q, \; \operatorname{tr}(Q) = 2k -n\}. \] As with the other two models, we show that differential geometric objects and operations -- tangent vector, metric, normal vector, exponential map, geodesic, parallel transport, gradient, Hessian, etc -- have closed-form analytic expressions that are computable with standard numerical linear algebra. In the proposed model, these expressions are considerably simpler, a result of representing $\operatorname{Gr}(k,n)$ as a linear section of a compact matrix Lie group $\operatorname{O}(n)$, and can be computed with at most one QR decomposition and one exponential of a special skew-symmetric matrix that takes only $O(nk(n-k))$ time. In particular, we completely avoid eigen- and singular value decompositions in our steepest descent, conjugate gradient, quasi-Newton, and Newton methods for the Grassmannian.

Practical gradient and conjugate gradient methods on flag manifolds

Convergence Analysis of Gradient Algorithms on Riemannian Manifolds Without Curvature Constraints and Application to Riemannian Mass

New vector transport operators extending a Riemannian CG algorithm to generalized Stiefel manifold with low-rank applications

Vector Transport Free Riemannian LBFGS for Optimization on Symmetric Positive Definite Matrix Manifolds

Linear Convergence of Subgradient Algorithm for Convex Feasibility on Riemannian Manifolds

Convergence analysis of the transformed gradient projection algorithms on compact matrix manifolds

Subgradient Projection Algorithms for Convex Feasibility on Riemannian Manifolds with Lower Bounded Curvatures

Gauss–Newton Method for Convex Composite Optimizations on Riemannian Manifolds

Convergence of Inexact Steepest Descent Algorithm for Multiobjective Optimizations on Riemannian Manifolds Without Curvature Constraints

Convergence analysis of inexact proximal point algorithms on Hadamard manifolds

Simpler Grassmannian optimization

The conjugate gradient method with various viewpoints

Riemannian Proximal Gradient Methods (extended Version)

Extragradient Type Methods for Riemannian Variational Inequality Problems

Optimization on product manifolds under a preconditioned metric

Alternative extension of the Hager–Zhang conjugate gradient method for vector optimization

The Proximal Gradient Method for Composite Optimization Problems on Riemannian Manifolds

On the extension of the Hager–Zhang conjugate gradient method for vector optimization

Nonsmooth Optimization over the Stiefel Manifold and Beyond: Proximal Gradient Method and Recent Variants

Riemannian proximal gradient methods