Abstract:
The condition number of a given mathematical problem is often related to
the reciprocal of its distance from the set of ill - conditioned problems. Such a property is
proved here for linear - quadratic convex optimization problems in the infinite - dimensional
setting. A uniform version of such theorem is obtained for suitably equi - bounded classes of
optimization problems. An application to the conditioning of a Ritz method is presented.
Keywords:
Condition numbers, linear-quadratic optimization, condition number theorem.
MSC:
90
|