A New Class of Feasible Direction Methods

摘要

A class of feasible direction methods based on two characterizations of optimality in convex programming is introduced. These methods are capable of generating all feasible directions, including those along the boundary of the feasible set.