Separable Augmented Lagrangian Algorithm with Multidimensional Scaling for Monotropic Programming
We analyze a new decomposition approach for convex structured programs based on augmented Lagrangian functions with multiple scaling parameters. We obtain global convergence results with weak hypotheses. Numerical results are presented on a class of multicommodity flow problems; empirical choices of the scaling parameters updates are discussed.
KeywordsAugmented Lagrangians monotropic programming multicommodity flows decomposition
Unable to display preview. Download preview PDF.