A feasible filter method for the nearest low-rank correlation matrix problem (Q494667): Difference between revisions

The paper presents a new easily computed constrained-preserving update scheme which can be viewed as a generalization of the Cayley transform, a classical retraction on the Stiefel manifold. A simple filter mechanism for unconstrained optimization is used in the given feasible method to promote global convergence to first-order critical points. Numerical results are given.