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.

Newton Geodesic Optimization on Special Linear Group

Kantorovich'S Theorems for Newton'S Method for Mappings and Optimization Problems on Lie Groups

Gauss–Newton Method for Convex Composite Optimizations on Riemannian Manifolds

Smale’s point estimate theory for Newton’s method on Lie groups

Newton's method on two step nilpotent lie groups: convergence criteria and the extended smale point estimate theory

Newton's Method on Riemannian Manifolds: Smale's Point Estimate Theory under the Γ-Condition

Optimality conditions for the nonlinear programming problems on Riemannian manifolds

EXTENDED NEWTON’S METHOD FOR MAPPINGS ON RIEMANNIAN MANIFOLDS WITH VALUES IN A CONE

Subspace Newton method for sparse group optimization problem

Critical Landscape Topology for Optimization on the Symplectic Group

Simpler Grassmannian optimization

Adaptive Quadratically Regularized Newton Method for Riemannian Optimization

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

Newton's Method for Sections on Riemannian Manifolds: Generalized Covariant Α-Theory

Geodesics, curvature, and conjugate points on Lie groups

A Geometric Perspective on Fusing Gaussian Distributions on Lie Groups

A Riemannian Proximal Newton-CG Method

Local Convergence of Newton's Method on the Heisenberg Group.

The SO(3) and SE(3) Lie Algebras of Rigid Body Rotations and Motions and their Application to Discrete Integration, Gradient Descent Optimization, and State Estimation

Symplectic Stiefel manifold: tractable metrics, second-order geometry and Newton's methods

Einstein Lie groups, geodesic orbit manifolds and regular Lie subgroups