Two general methods for computing saddle points with applications for decomposing convex programming problems (Q1069674)

Two new methods are presented for computing saddle points in finite- dimensional spaces. One of the methods is a modification of the column method given by Dantzig and Wolfe, obtained by regularizing the Lagrangian. The other method is a symmetric modification of the above. The principal application of the methods presented here is to give new decomposition algorithms for solving separable convex programming problems.