An inertia theorem for symmetric matrices and its application to nonlinear programming (Q1068908): Difference between revisions

The authors study a relationship between the inertia (i.e. the ordered triple of positive, negative, and zero eigenvalues) of a real symmetric matrix and the inertia of its restriction to a subspace. The result is then applied to the partitioned matrix \(A=(A_{ij})\) with \(A_{11}=H\), \(A_{12}=J^ T\), \(A_{21}=J\), \(A_{22}=0\), where H is the Hessian of the Lagrangian and J is the Jacobian of the constraints in nonlinear programming. It is shown that the inertia of such a matrix A depends only on the inertia of the projection of H onto the null-space of J. Hence it is possible to describe the Karush-Kuhn-Tucker points by the inertia of A rather than H.