A regularized decomposition method for minimizing a sum of polyhedral functions

From MaRDI portal





Motivated by the two-stage stochastic programming problem with recourse for the case of a discrete distribution of the involved random elements, the author proposes an algorithm for the solution of nonlinear programming problems where the objective function to be minimized is the sum of many convex piecewise-linear functions, the constraints are linear and convex piecewise-linear functions, too. The method generates two sequences: main iterates \(\{x^ k\}\) and trial points \(\{y^ k\}\). At each iteration each of the functions are approximated from below by a pointwise maximum of a finite collection of linear functions and an approximate problem is constructed to find the next trial point. The objective of the approximate problem contains a quadratic term to stabilize it and to keep the number of the elements in the finite collection limited. The approximate problem is solved in each iteration by a special algorithm based on an active set strategy. Convergence analysis is provided together with computational experiences.



Cites work


Cited in
(99)








This page was built for publication: A regularized decomposition method for minimizing a sum of polyhedral functions

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3735489)