Low rank differential equations for Hamiltonian matrix nearness problems (Q2514247)

The authors propose and study algorithms for two important optimization problems related to finding nearest Hamiltonian matrices with imaginary eigenvalues, to two specific classes of matrices. Their algorithms consist of a two-lewel procedure, where on the inner level an extremizer for a fixed perturbation size is determined, following the differential equation into a stationary point, and then the outer level to optimize over this size.