Equivalence of two nondegeneracy conditions for semidefinite programs (Q2483041)

In semidefinite programming, a broad generalization of linear programming to the convex cone of positive semidefinite matrices, degenerate problems are typically difficult to solve: algorithms show slow convergence near the optimum, and computed solutions may be highly sensitive to perturbations. It is thus relevant to derive conditions for non-degeneracy. In this paper, the authors show that two previously available non-degeneracy conditions are actually equivalent.