This solves LPs using the predictor-corrector method due to Mehrotra. More...
#include <ClpPredictorCorrector.hpp>
Public Member Functions | |
Description of algorithm | |
int | solve () |
Primal Dual Predictor Corrector algorithm. | |
Functions used in algorithm | |
CoinWorkDouble | findStepLength (int phase) |
findStepLength. | |
CoinWorkDouble | findDirectionVector (const int phase) |
findDirectionVector. | |
int | createSolution () |
createSolution. Creates solution from scratch (- code if no memory) | |
CoinWorkDouble | complementarityGap (int &numberComplementarityPairs, int &numberComplementarityItems, const int phase) |
complementarityGap. Computes gap | |
void | setupForSolve (const int phase) |
setupForSolve. | |
void | solveSystem (CoinWorkDouble *region1, CoinWorkDouble *region2, const CoinWorkDouble *region1In, const CoinWorkDouble *region2In, const CoinWorkDouble *saveRegion1, const CoinWorkDouble *saveRegion2, bool gentleRefine) |
Does solve. | |
bool | checkGoodMove (const bool doCorrector, CoinWorkDouble &bestNextGap, bool allowIncreasingGap) |
sees if looks plausible change in complementarity | |
bool | checkGoodMove2 (CoinWorkDouble move, CoinWorkDouble &bestNextGap, bool allowIncreasingGap) |
: checks for one step size | |
int | updateSolution (CoinWorkDouble nextGap) |
updateSolution. Updates solution at end of iteration | |
CoinWorkDouble | affineProduct () |
Save info on products of affine deltaT*deltaW and deltaS*deltaZ. | |
void | debugMove (int phase, CoinWorkDouble primalStep, CoinWorkDouble dualStep) |
See exactly what would happen given current deltas. |
This solves LPs using the predictor-corrector method due to Mehrotra.
It also uses multiple centrality corrections as in Gondzio.
See; S. Mehrotra, "On the implementation of a primal-dual interior point method", SIAM Journal on optimization, 2 (1992) J. Gondzio, "Multiple centraility corrections in a primal-dual method for linear programming", Computational Optimization and Applications",6 (1996)
It is rather basic as Interior point is not my speciality
It inherits from ClpInterior. It has no data of its own and is never created - only cast from a ClpInterior object at algorithm time.
It can also solve QPs
Definition at line 37 of file ClpPredictorCorrector.hpp.
int ClpPredictorCorrector::solve | ( | ) |
Primal Dual Predictor Corrector algorithm.
Method
Big TODO
CoinWorkDouble ClpPredictorCorrector::findStepLength | ( | int | phase | ) |
findStepLength.
CoinWorkDouble ClpPredictorCorrector::findDirectionVector | ( | const int | phase | ) |
findDirectionVector.
createSolution. Creates solution from scratch (- code if no memory)
CoinWorkDouble ClpPredictorCorrector::complementarityGap | ( | int & | numberComplementarityPairs, |
int & | numberComplementarityItems, | ||
const int | phase | ||
) |
complementarityGap. Computes gap
void ClpPredictorCorrector::setupForSolve | ( | const int | phase | ) |
setupForSolve.
void ClpPredictorCorrector::solveSystem | ( | CoinWorkDouble * | region1, |
CoinWorkDouble * | region2, | ||
const CoinWorkDouble * | region1In, | ||
const CoinWorkDouble * | region2In, | ||
const CoinWorkDouble * | saveRegion1, | ||
const CoinWorkDouble * | saveRegion2, | ||
bool | gentleRefine | ||
) |
Does solve.
region1 is for deltaX (columns+rows), region2 for deltaPi (rows)
bool ClpPredictorCorrector::checkGoodMove | ( | const bool | doCorrector, |
CoinWorkDouble & | bestNextGap, | ||
bool | allowIncreasingGap | ||
) |
sees if looks plausible change in complementarity
bool ClpPredictorCorrector::checkGoodMove2 | ( | CoinWorkDouble | move, |
CoinWorkDouble & | bestNextGap, | ||
bool | allowIncreasingGap | ||
) |
: checks for one step size
int ClpPredictorCorrector::updateSolution | ( | CoinWorkDouble | nextGap | ) |
updateSolution. Updates solution at end of iteration
CoinWorkDouble ClpPredictorCorrector::affineProduct | ( | ) |
Save info on products of affine deltaT*deltaW and deltaS*deltaZ.
void ClpPredictorCorrector::debugMove | ( | int | phase, |
CoinWorkDouble | primalStep, | ||
CoinWorkDouble | dualStep | ||
) |
See exactly what would happen given current deltas.