Active-Set Methods for Quadratic Programming: Where are We Now?

Recent research on quadratic programming (QP) methods has tended to focus on special techniques for bound constraints, on strategies for making multiple changes in the working set, and on interior/barrier-function approaches. Nonetheless, the state of the art in what might be termed 'standard' active-set QP methods is not always clear. This talk summarizes selected aspects of theory and implementation for active-set QP methods, with emphasis on techniques for sparse problems based on the Schur complement. Stress will be placed on theoretical similarities and practical differences in widely used methods.