Abstract
Difficulties experienced with slow terminal convergence near the optimum when using the supporting hyperplane algorithm are shown to be slightly alleviated by the use of a quadratic approximation that is very simple to use. Some comments from experience with the use of this technique are given.