Couenne::CouenneFeasPump Class Reference

An implementation of the Feasibility pump that uses linearization and Ipopt to find the two sequences of points. More...

#include <CouenneFeasPump.hpp>

Collaboration diagram for Couenne::CouenneFeasPump:
Collaboration graph
[legend]

List of all members.

Public Types

enum  fpCompDistIntType { FP_DIST_INT, FP_DIST_ALL, FP_DIST_POST }
enum  fpCutPlane { FP_CUT_NONE, FP_CUT_INTEGRATED, FP_CUT_EXTERNAL, FP_CUT_POST }
enum  fpTabuMgtPolicy { FP_TABU_NONE, FP_TABU_POOL, FP_TABU_PERTURB, FP_TABU_CUT }

Public Member Functions

 CouenneFeasPump (CouenneProblem *couenne=NULL, CouenneCutGenerator *cg=NULL, Ipopt::SmartPtr< Ipopt::OptionsList > options=NULL)
 Constructor with (optional) MINLP pointer.
 CouenneFeasPump (const CouenneFeasPump &other)
 Copy constructor.
virtual ~CouenneFeasPump ()
 Destructor.
virtual CbcHeuristic * clone () const
 Clone.
CouenneFeasPumpoperator= (const CouenneFeasPump &rhs)
 Assignment operator.
virtual void resetModel (CbcModel *model)
 Does nothing, but necessary as CbcHeuristic declares it pure virtual.
virtual int solution (double &objectiveValue, double *newSolution)
 Run heuristic, return 1 if a better solution than the one passed is found and 0 otherwise.
void setNumberSolvePerLevel (int value)
 set number of nlp's solved for each given level of the tree
virtual CouNumber solveMILP (const CouNumber *nSol, CouNumber *&iSol, int niter, int *nsuciter)
 find integer (possibly NLP-infeasible) point isol closest (according to the l-1 norm of the hessian) to the current NLP-feasible (but fractional) solution nsol
virtual CouNumber solveNLP (const CouNumber *nSol, CouNumber *&iSol)
 obtain solution to NLP
expressionupdateNLPObj (const double *)
 set new expression as the NLP objective function using argument as point to minimize distance from.
bool fixIntVariables (const double *sol)
 admits a (possibly fractional) solution and fixes the integer components in the nonlinear problem for later re-solve.
double findSolution (const double *nSol, double *&sol, int niter, int *nsuciter)
 find feasible solution (called by solveMILP ())
void init_MILP ()
 initialize all solvers at the first call, where the initial MILP is built
void initIpoptApp ()
 Common code for initializing non-smartptr ipopt application.
CouenneProblemProblem () const
 return pointer to problem
enum fpCompDistIntType compDistInt () const
 return type of MILP solved
double multDistNLP () const
 Return Weights in computing distance, in both MILP and NLP (must sum up to 1 for MILP and for NLP):.
double multHessNLP () const
 weight of Hessian in NLP
double multObjFNLP () const
 weight of objective in NLP
double multDistMILP () const
 weight of distance in MILP
double multHessMILP () const
 weight of Hessian in MILP
double multObjFMILP () const
 weight of objective in MILP
CouenneTNLPnlp () const
 return NLP
int & nCalls ()
 return number of calls (can be changed)
int milpPhase (double *nSol, double *iSol)
 MILP phase of the FP.
int nlpPhase (double *iSol, double *nSol)
 NLP phase of the FP.

Static Public Member Functions

static void registerOptions (Ipopt::SmartPtr< Bonmin::RegisteredOptions >)
 initialize options to be read later

Private Attributes

CouenneProblemproblem_
 Couenne representation of the problem.
CouenneCutGeneratorcouenneCG_
 CouenneCutGenerator for linearization cuts.
CouenneTNLPnlp_
 Continuous relaxation of the problem, with an interface for Ipopt only.
Ipopt::IpoptApplication * app_
 Ipopt Application pointer for solving NLPs.
OsiSolverInterface * milp_
 MILP relaxation of the MINLP (used to find integer, non-NLP-feasible solutions).
OsiSolverInterface * postlp_
 LP relaxation of the MINLP used when fixing integer variables (used for compDistInt_ in FP_DIST_POST and possibly FP_DIST_INT).
CouenneFPpoolpool_
 Pool of solutions.
std::set< CouenneFPsolution,
compareSol
tabuPool_
 Solutions to avoid.
int * match_
 matching between reformulation's variables and L-1 norm variables
int numberSolvePerLevel_
 Number of NLPs solved for each given level of the tree.
double multDistNLP_
 Weights in computing distance, in both MILP and NLP (must sum up to 1 for MILP and for NLP):.
double multHessNLP_
 weight of Hessian in NLP
double multObjFNLP_
 weight of objective in NLP
double multDistMILP_
 weight of distance in MILP
double multHessMILP_
 weight of Hessian in MILP
double multObjFMILP_
 weight of objective in MILP
enum fpCompDistIntType compDistInt_
 Compute distance from integer variables only, not all variables.
enum fpCutPlane milpCuttingPlane_
 Separate convexification cuts during or after MILP.
int nSepRounds_
 Number of separation rounds for MILP convexification cuts.
int maxIter_
 Maximum iterations per call.
bool useSCIP_
 Use SCIP instead of Cbc for solving MILPs.
int milpMethod_
 Which SCIP MILP method to use.
enum fpTabuMgtPolicy tabuMgt_
 Tabu management policy: none, use from pool, random perturbation of current solution.
int nCalls_
 How often should it be called.
double fadeMult_
 decrease factor for MILP/NLP multipliers of distance/Hessian/objective

Detailed Description

An implementation of the Feasibility pump that uses linearization and Ipopt to find the two sequences of points.

Definition at line 57 of file CouenneFeasPump.hpp.


Member Enumeration Documentation

Enumerator:
FP_DIST_INT 
FP_DIST_ALL 
FP_DIST_POST 

Definition at line 61 of file CouenneFeasPump.hpp.

Enumerator:
FP_CUT_NONE 
FP_CUT_INTEGRATED 
FP_CUT_EXTERNAL 
FP_CUT_POST 

Definition at line 62 of file CouenneFeasPump.hpp.

Enumerator:
FP_TABU_NONE 
FP_TABU_POOL 
FP_TABU_PERTURB 
FP_TABU_CUT 

Definition at line 63 of file CouenneFeasPump.hpp.


Constructor & Destructor Documentation

CouenneFeasPump::CouenneFeasPump ( CouenneProblem couenne = NULL,
CouenneCutGenerator cg = NULL,
Ipopt::SmartPtr< Ipopt::OptionsList >  options = NULL 
)

Constructor with (optional) MINLP pointer.

Definition at line 58 of file CouenneFeasPumpConstructors.cpp.

CouenneFeasPump::CouenneFeasPump ( const CouenneFeasPump other  ) 

Copy constructor.

Definition at line 161 of file CouenneFeasPumpConstructors.cpp.

CouenneFeasPump::~CouenneFeasPump (  )  [virtual]

Destructor.

Definition at line 221 of file CouenneFeasPumpConstructors.cpp.


Member Function Documentation

CbcHeuristic * CouenneFeasPump::clone (  )  const [virtual]

Clone.

Definition at line 166 of file CouenneFeasPumpConstructors.cpp.

CouenneFeasPump & CouenneFeasPump::operator= ( const CouenneFeasPump rhs  ) 

Assignment operator.

Definition at line 171 of file CouenneFeasPumpConstructors.cpp.

virtual void Couenne::CouenneFeasPump::resetModel ( CbcModel *  model  )  [inline, virtual]

Does nothing, but necessary as CbcHeuristic declares it pure virtual.

Definition at line 83 of file CouenneFeasPump.hpp.

int CouenneFeasPump::solution ( double &  objectiveValue,
double *  newSolution 
) [virtual]

Run heuristic, return 1 if a better solution than the one passed is found and 0 otherwise.

objectiveValue Best known solution in input and value of solution found in output

newSolution Solution found by heuristic.

Definition at line 39 of file CouenneFeasPump.cpp.

void Couenne::CouenneFeasPump::setNumberSolvePerLevel ( int  value  )  [inline]

set number of nlp's solved for each given level of the tree

Definition at line 95 of file CouenneFeasPump.hpp.

CouNumber CouenneFeasPump::solveMILP ( const CouNumber nSol,
CouNumber *&  iSol,
int  niter,
int *  nsuciter 
) [virtual]

find integer (possibly NLP-infeasible) point isol closest (according to the l-1 norm of the hessian) to the current NLP-feasible (but fractional) solution nsol

find integer (possibly NLP-infeasible) point isol closest (according to the l-1 norm of the Hessian) to the current NLP-feasible (but fractional) solution nsol

Definition at line 53 of file CouenneFPSolveMILP.cpp.

CouNumber CouenneFeasPump::solveNLP ( const CouNumber nSol,
CouNumber *&  iSol 
) [virtual]

obtain solution to NLP

obtain continuous (if fractional) solution

Definition at line 27 of file CouenneFPSolveNLP.cpp.

expression * CouenneFeasPump::updateNLPObj ( const double *  iSol  ) 

set new expression as the NLP objective function using argument as point to minimize distance from.

Set new expression as the NLP objective function using argument as point to minimize distance from.

Return new objective function

Return new objective function.

Definition at line 235 of file CouenneFeasPumpConstructors.cpp.

bool CouenneFeasPump::fixIntVariables ( const double *  sol  ) 

admits a (possibly fractional) solution and fixes the integer components in the nonlinear problem for later re-solve.

Reads a (possibly fractional) solution and fixes the integer components in the nonlinear problem for later re-solve.

Returns false if restriction infeasible, true otherwise

Definition at line 428 of file CouenneFeasPumpConstructors.cpp.

void CouenneFeasPump::registerOptions ( Ipopt::SmartPtr< Bonmin::RegisteredOptions roptions  )  [static]

initialize options to be read later

initialize options

Definition at line 471 of file CouenneFeasPumpConstructors.cpp.

double CouenneFeasPump::findSolution ( const double *  nSol,
double *&  sol,
int  niter,
int *  nsuciter 
)

find feasible solution (called by solveMILP ())

find a feasible or optimal solution of MILP

As found on the notes, these methods can be used, from the most expensive and accurate (exact) method to a cheap, inexact one:

1. Solve a MILP relaxation with Manhattan distance as objective 2. Partially solve the MILP with emphasis on good solutions 3. Apply RENS to 1 4. Use Objective FP 2.0 for MILPs 5. round-and-propagate 6. choose from pool, see 4 7. random perturbation

solve MILP

Definition at line 27 of file CouenneFPFindSolution.cpp.

void CouenneFeasPump::init_MILP (  ) 

initialize all solvers at the first call, where the initial MILP is built

initialize MILP solvers if needed

Definition at line 108 of file CouenneFPFindSolution.cpp.

void CouenneFeasPump::initIpoptApp (  ) 

Common code for initializing non-smartptr ipopt application.

Definition at line 31 of file CouenneFeasPumpConstructors.cpp.

CouenneProblem* Couenne::CouenneFeasPump::Problem (  )  const [inline]

return pointer to problem

Definition at line 135 of file CouenneFeasPump.hpp.

enum fpCompDistIntType Couenne::CouenneFeasPump::compDistInt (  )  const [inline]

return type of MILP solved

Definition at line 139 of file CouenneFeasPump.hpp.

double Couenne::CouenneFeasPump::multDistNLP (  )  const [inline]

Return Weights in computing distance, in both MILP and NLP (must sum up to 1 for MILP and for NLP):.

weight of distance in NLP

Definition at line 145 of file CouenneFeasPump.hpp.

double Couenne::CouenneFeasPump::multHessNLP (  )  const [inline]

weight of Hessian in NLP

Definition at line 146 of file CouenneFeasPump.hpp.

double Couenne::CouenneFeasPump::multObjFNLP (  )  const [inline]

weight of objective in NLP

Definition at line 147 of file CouenneFeasPump.hpp.

double Couenne::CouenneFeasPump::multDistMILP (  )  const [inline]

weight of distance in MILP

Definition at line 149 of file CouenneFeasPump.hpp.

double Couenne::CouenneFeasPump::multHessMILP (  )  const [inline]

weight of Hessian in MILP

Definition at line 150 of file CouenneFeasPump.hpp.

double Couenne::CouenneFeasPump::multObjFMILP (  )  const [inline]

weight of objective in MILP

Definition at line 151 of file CouenneFeasPump.hpp.

CouenneTNLP* Couenne::CouenneFeasPump::nlp (  )  const [inline]

return NLP

Definition at line 154 of file CouenneFeasPump.hpp.

int& Couenne::CouenneFeasPump::nCalls (  )  [inline]

return number of calls (can be changed)

Definition at line 158 of file CouenneFeasPump.hpp.

int Couenne::CouenneFeasPump::milpPhase ( double *  nSol,
double *  iSol 
)

MILP phase of the FP.

int Couenne::CouenneFeasPump::nlpPhase ( double *  iSol,
double *  nSol 
)

NLP phase of the FP.


Member Data Documentation

Couenne representation of the problem.

Definition at line 179 of file CouenneFeasPump.hpp.

CouenneCutGenerator for linearization cuts.

Definition at line 182 of file CouenneFeasPump.hpp.

Continuous relaxation of the problem, with an interface for Ipopt only.

Definition at line 193 of file CouenneFeasPump.hpp.

Ipopt::IpoptApplication* Couenne::CouenneFeasPump::app_ [private]

Ipopt Application pointer for solving NLPs.

Definition at line 196 of file CouenneFeasPump.hpp.

OsiSolverInterface* Couenne::CouenneFeasPump::milp_ [private]

MILP relaxation of the MINLP (used to find integer, non-NLP-feasible solutions).

Definition at line 200 of file CouenneFeasPump.hpp.

OsiSolverInterface* Couenne::CouenneFeasPump::postlp_ [private]

LP relaxation of the MINLP used when fixing integer variables (used for compDistInt_ in FP_DIST_POST and possibly FP_DIST_INT).

Definition at line 205 of file CouenneFeasPump.hpp.

Pool of solutions.

Definition at line 208 of file CouenneFeasPump.hpp.

Solutions to avoid.

Definition at line 211 of file CouenneFeasPump.hpp.

matching between reformulation's variables and L-1 norm variables

Definition at line 214 of file CouenneFeasPump.hpp.

Number of NLPs solved for each given level of the tree.

Definition at line 221 of file CouenneFeasPump.hpp.

Weights in computing distance, in both MILP and NLP (must sum up to 1 for MILP and for NLP):.

weight of distance in NLP

Definition at line 226 of file CouenneFeasPump.hpp.

weight of Hessian in NLP

Definition at line 227 of file CouenneFeasPump.hpp.

weight of objective in NLP

Definition at line 228 of file CouenneFeasPump.hpp.

weight of distance in MILP

Definition at line 230 of file CouenneFeasPump.hpp.

weight of Hessian in MILP

Definition at line 231 of file CouenneFeasPump.hpp.

weight of objective in MILP

Definition at line 232 of file CouenneFeasPump.hpp.

Compute distance from integer variables only, not all variables.

Definition at line 235 of file CouenneFeasPump.hpp.

Separate convexification cuts during or after MILP.

Definition at line 238 of file CouenneFeasPump.hpp.

Number of separation rounds for MILP convexification cuts.

Definition at line 241 of file CouenneFeasPump.hpp.

Maximum iterations per call.

Definition at line 244 of file CouenneFeasPump.hpp.

Use SCIP instead of Cbc for solving MILPs.

Definition at line 247 of file CouenneFeasPump.hpp.

Which SCIP MILP method to use.

Definition at line 250 of file CouenneFeasPump.hpp.

Tabu management policy: none, use from pool, random perturbation of current solution.

Definition at line 253 of file CouenneFeasPump.hpp.

How often should it be called.

Definition at line 256 of file CouenneFeasPump.hpp.

decrease factor for MILP/NLP multipliers of distance/Hessian/objective

Definition at line 259 of file CouenneFeasPump.hpp.


The documentation for this class was generated from the following files:

Generated on 5 Jul 2015 by  doxygen 1.6.1