Definition of the SubGrad class, which implements the NDOSolver interface for NonDifferentiable Optimization Solvers, as described in NDOSlver.h, using a (deflected, projected, incremental) subgradient-type algorithm. More...

#include "NDOSlver.h"
#include "Deflection.h"
#include "Stepsize.h"
#include "CQKnPClass.h"
#include <vector>
#include <algorithm>

Classes
class	SubGrad
	The `SubGrad` class implements the NDOSolver interface for NonDifferentiable Optimization Solvers [see NDOSlver.h], using a unified subgradient-type algorithm as described in: More...

Macros
#define	LOG_SG 1
	If LOG_SG > 0, the SubGrad class produces a log of its activities on the ostream object and at the "level of verbosity" set with the method SetNDOLog() [see below]. More...

Detailed Description

Definition of the SubGrad class, which implements the NDOSolver interface for NonDifferentiable Optimization Solvers, as described in NDOSlver.h, using a (deflected, projected, incremental) subgradient-type algorithm.

The user is assumed to be familiar with the algorithm: refer to

A. Frangioni, E. Gorgone, B. Gendron. On the Computational Efficiency of Subgradient Methods: A Case Study in Combinatorial Optimization . Technical Report CIRRELT- 2015-41, Interuniversity Research Centre on Enterprise Networks, Logistics and Transportation (CIRRELT), Montreal, Canada, 2015

Version: 1.00

Date: 27 - 10 - 2015

Author: Antonio Frangioni
Operations Research Group
Department of Informatics
University of Pisa; Enrico Gorgone
Graphs and Mathematical Optimization (GOM) Group
Department of Informatics
Free University of Brussels

The class requires that the function to be minimized be available under the form of a FiOracle object, as described in FiOracle.h.

The class is parametric over the type of both stepsize and direction used: it just relies over the objects of classes Stepsize and Deflection.

Classes

Macros

Detailed Description