Conditional quadratic semidefinite programming: examples and methods (Q489109)

By conditional semidefinite programming the author means a linear programming problem in the cone of positive semidefinite matrices projected in a lower dimensional subspace. The author argue that it is a relevant class of convex optimization problems by enumerating a list of examples from the literature (matrix completion, matrix approximation, Euclidean distance matrices), providing a new example of calibration of correlation matrix from mathematical finance (in section 5), and describing a semismooth quasi-Newton-conjugate-gradient algorithm with favorable convergence properties.