//#############################################################################
void
MibSHeuristic::lowerObjHeuristic()
{
/*
optimize wrt to lower-level objective
over current feasible lp feasible region
*/
MibSModel * model = MibSModel_;
OsiSolverInterface * oSolver = model->getSolver();
//OsiSolverInterface * hSolver = new OsiCbcSolverInterface();
OsiSolverInterface* hSolver = new OsiSymSolverInterface();
double objSense(model->getLowerObjSense());
int lCols(model->getLowerDim());
int uCols(model->getUpperDim());
int * lColIndices = model->getLowerColInd();
int * uColIndices = model->getUpperColInd();
double * lObjCoeffs = model->getLowerObjCoeffs();
//int tCols(lCols + uCols);
int tCols(oSolver->getNumCols());
//assert(tCols == oSolver->getNumCols());
hSolver->loadProblem(*oSolver->getMatrixByCol(),
oSolver->getColLower(), oSolver->getColUpper(),
oSolver->getObjCoefficients(),
oSolver->getRowLower(), oSolver->getRowUpper());
int j(0);
for(j = 0; j < tCols; j++){
if(oSolver->isInteger(j))
hSolver->setInteger(j);
}
double * nObjCoeffs = new double[tCols];
int i(0), index(0);
CoinZeroN(nObjCoeffs, tCols);
for(i = 0; i < lCols; i++){
index = lColIndices[i];
nObjCoeffs[index] = lObjCoeffs[i];
}
//MibS objective sense is the opposite of OSI's!
hSolver->setObjSense(objSense);
hSolver->setObjective(nObjCoeffs);
//double cutoff(model->getCutoff());
double cutoff(model->getKnowledgeBroker()->getIncumbentValue());
if(model->getNumSolutions()){
CoinPackedVector objCon;
//double rhs(cutoff * objSense);
//double smlTol(1.0);
double rhs(cutoff);
for(i = 0; i < tCols; i++){
objCon.insert(i, oSolver->getObjCoefficients()[i]
* oSolver->getObjSense());
}
hSolver->addRow(objCon, - hSolver->getInfinity(), rhs);
}
if(0)
hSolver->writeLp("lobjheurstic");
if(0){
dynamic_cast<OsiCbcSolverInterface *>
(hSolver)->getModelPtr()->messageHandler()->setLogLevel(0);
}
else{
dynamic_cast<OsiSymSolverInterface *>
(hSolver)->setSymParam("prep_level", -1);
dynamic_cast<OsiSymSolverInterface *>
(hSolver)->setSymParam("verbosity", -2);
dynamic_cast<OsiSymSolverInterface *>
(hSolver)->setSymParam("max_active_nodes", 1);
}
hSolver->branchAndBound();
if(hSolver->isProvenOptimal()){
double upperObjVal(0.0);
/*****************NEW ******************/
MibSSolution *mibSol = NULL;
OsiSolverInterface * lSolver = model->bS_->setUpModel(hSolver, true);
if(0){
lSolver->writeLp("tmp");
}
if(0){
dynamic_cast<OsiCbcSolverInterface *>
(lSolver)->getModelPtr()->messageHandler()->setLogLevel(0);
}
else{
dynamic_cast<OsiSymSolverInterface *>
(lSolver)->setSymParam("prep_level", -1);
dynamic_cast<OsiSymSolverInterface *>
(lSolver)->setSymParam("verbosity", -2);
dynamic_cast<OsiSymSolverInterface *>
(lSolver)->setSymParam("max_active_nodes", 1);
}
lSolver->branchAndBound();
if (lSolver->isProvenOptimal()){
const double * sol = hSolver->getColSolution();
double objVal(lSolver->getObjValue() * objSense);
double etol(etol_);
double lowerObj = getLowerObj(sol, objSense);
double * optUpperSolutionOrd = new double[uCols];
double * optLowerSolutionOrd = new double[lCols];
CoinZeroN(optUpperSolutionOrd, uCols);
CoinZeroN(optLowerSolutionOrd, lCols);
if(fabs(objVal - lowerObj) < etol){
/** Current solution is bilevel feasible **/
for(i = 0; i < tCols; i++)
upperObjVal +=
hSolver->getColSolution()[i] * oSolver->getObjCoefficients()[i];
mibSol = new MibSSolution(hSolver->getNumCols(),
hSolver->getColSolution(),
upperObjVal,
model);
model->storeSolution(BlisSolutionTypeHeuristic, mibSol);
mibSol = NULL;
}
else{
/* solution is not bilevel feasible, create one that is */
const double * uSol = hSolver->getColSolution();
const double * lSol = lSolver->getColSolution();
int numElements(hSolver->getNumCols());
int i(0), pos(0), index(0);
double * lpSolution = new double[numElements];
double upperObj(0.0);
//FIXME: problem is still here. indices may be wrong.
//also is all this necessary, or can we just paste together uSol and lSol?
//this may be an old comment
for(i = 0; i < numElements; i++){
pos = model->bS_->binarySearch(0, lCols - 1, i, lColIndices);
if(pos < 0){
pos = model->bS_->binarySearch(0, uCols - 1, i, uColIndices);
if (pos >= 0){
optUpperSolutionOrd[pos] = uSol[i];
}
}
else{
optLowerSolutionOrd[pos] = lSol[pos];
}
}
for(i = 0; i < uCols; i++){
index = uColIndices[i];
lpSolution[index] = optUpperSolutionOrd[i];
upperObj +=
optUpperSolutionOrd[i] * oSolver->getObjCoefficients()[index];
}
for(i = 0; i < lCols; i++){
index = lColIndices[i];
lpSolution[index] = optLowerSolutionOrd[i];
upperObj +=
optLowerSolutionOrd[i] * oSolver->getObjCoefficients()[index];
}
if(model->checkUpperFeasibility(lpSolution)){
mibSol = new MibSSolution(hSolver->getNumCols(),
lpSolution,
upperObj * oSolver->getObjSense(),
model);
model->storeSolution(BlisSolutionTypeHeuristic, mibSol);
mibSol = NULL;
}
delete [] lpSolution;
}
}
delete lSolver;
}
delete hSolver;
}
//#############################################################################
mcSol
MibSHeuristic::solveSubproblem(double beta)
{
/*
optimize wrt to weighted upper-level objective
over current feasible lp feasible region
*/
MibSModel * model = MibSModel_;
OsiSolverInterface * oSolver = model->getSolver();
//OsiSolverInterface * sSolver = new OsiCbcSolverInterface();
OsiSolverInterface* sSolver = new OsiSymSolverInterface();
//sSolver = oSolver->clone();
//OsiSolverInterface * sSolver = tmpSolver;
//OsiSolverInterface * tmpSolver = new OsiSolverInterface(oSolver);
double uObjSense(oSolver->getObjSense());
double lObjSense(model->getLowerObjSense());
int lCols(model->getLowerDim());
int uCols(model->getUpperDim());
int * lColIndices = model->getLowerColInd();
int * uColIndices = model->getUpperColInd();
double * lObjCoeffs = model->getLowerObjCoeffs();
const double * uObjCoeffs = oSolver->getObjCoefficients();
double etol(etol_);
int tCols(uCols + lCols);
assert(tCols == oSolver->getNumCols());
sSolver->loadProblem(*oSolver->getMatrixByCol(),
oSolver->getColLower(), oSolver->getColUpper(),
oSolver->getObjCoefficients(),
oSolver->getRowLower(), oSolver->getRowUpper());
int j(0);
for(j = 0; j < tCols; j++){
if(oSolver->isInteger(j))
sSolver->setInteger(j);
}
double * nObjCoeffs = new double[tCols];
int i(0), index(0);
CoinZeroN(nObjCoeffs, tCols);
/* Multiply the UL columns of the UL objective by beta */
for(i = 0; i < uCols; i++){
index = uColIndices[i];
if(fabs(uObjCoeffs[index]) > etol)
nObjCoeffs[index] = beta * uObjCoeffs[index] * uObjSense;
else
nObjCoeffs[index] = 0.0;
}
/* Multiply the LL columns of the UL objective by beta */
for(i = 0; i < lCols; i++){
index = lColIndices[i];
if(fabs(uObjCoeffs[index]) > etol)
nObjCoeffs[index] = beta* uObjCoeffs[index] * uObjSense;
else
nObjCoeffs[index] = 0.0;
}
/* Add the LL columns of the LL objective multiplied by (1 - beta) */
for(i = 0; i < lCols; i++){
index = lColIndices[i];
if(fabs(lObjCoeffs[i]) > etol)
nObjCoeffs[index] += (1 - beta) * lObjCoeffs[i] * lObjSense;
}
sSolver->setObjective(nObjCoeffs);
//int i(0);
if(0){
for(i = 0; i < sSolver->getNumCols(); i++){
std::cout << "betaobj " << sSolver->getObjCoefficients()[i] << std::endl;
}
}
if(0){
sSolver->writeLp("afterbeta");
//sSolver->writeMps("afterbeta");
}
if(0){
for(i = 0; i < sSolver->getNumCols(); i++){
std::cout << "obj " << sSolver->getObjCoefficients()[i] << std::endl;
std::cout << "upper " << sSolver->getColUpper()[i] << std::endl;
std::cout << "lower " << sSolver->getColLower()[i] << std::endl;
}
}
if(0){
dynamic_cast<OsiCbcSolverInterface *>
(sSolver)->getModelPtr()->messageHandler()->setLogLevel(0);
}
else{
dynamic_cast<OsiSymSolverInterface *>
(sSolver)->setSymParam("prep_level", -1);
dynamic_cast<OsiSymSolverInterface *>
(sSolver)->setSymParam("verbosity", -2);
dynamic_cast<OsiSymSolverInterface *>
(sSolver)->setSymParam("max_active_nodes", 1);
}
//dynamic_cast<OsiSymSolverInterface *> (sSolver)->branchAndBound();
sSolver->branchAndBound();
if(sSolver->isProvenOptimal()){
if(0){
std::cout << "writing lp file." << std::endl;
sSolver->writeLp("afterbeta");
//sSolver->writeMps("afterbeta");
}
double upperObjVal(0.0);
double lowerObjVal(0.0);
for(i = 0; i < tCols; i++){
upperObjVal +=
sSolver->getColSolution()[i] * oSolver->getObjCoefficients()[i];
if(0){
std::cout << "sSolver->getColSolution()[" << i << "] :"
<< sSolver->getColSolution()[i] << std::endl;
}
}
lowerObjVal = getLowerObj(sSolver->getColSolution(), lObjSense);
if(beta == 1.0){
/*
fix upper-level objective to current value and
reoptimize wrt to lower-level objective
*/
//OsiSolverInterface * nSolver = new OsiCbcSolverInterface();
OsiSolverInterface * nSolver = new OsiSymSolverInterface();
nSolver->loadProblem(*oSolver->getMatrixByCol(),
oSolver->getColLower(), oSolver->getColUpper(),
oSolver->getObjCoefficients(),
oSolver->getRowLower(), oSolver->getRowUpper());
for(j = 0; j < tCols; j++){
if(oSolver->isInteger(j))
nSolver->setInteger(j);
}
CoinZeroN(nObjCoeffs, tCols);
for(i = 0; i < lCols; i++){
index = lColIndices[i];
nObjCoeffs[index] = lObjCoeffs[i] * lObjSense;
}
nSolver->setObjective(nObjCoeffs);
CoinPackedVector objCon;
for(i = 0; i < tCols; i++){
objCon.insert(i, uObjCoeffs[i] * uObjSense);
}
nSolver->addRow(objCon, upperObjVal, upperObjVal);
nSolver->writeLp("beta1");
if(0){
dynamic_cast<OsiCbcSolverInterface *>
(nSolver)->getModelPtr()->messageHandler()->setLogLevel(0);
}
else{
dynamic_cast<OsiSymSolverInterface *>
(nSolver)->setSymParam("prep_level", -1);
dynamic_cast<OsiSymSolverInterface *>
(nSolver)->setSymParam("verbosity", -2);
dynamic_cast<OsiSymSolverInterface *>
(nSolver)->setSymParam("max_active_nodes", 1);
}
nSolver->branchAndBound();
double * colsol = new double[tCols];
if(nSolver->isProvenOptimal()){
lowerObjVal = nSolver->getObjValue();
CoinCopyN(nSolver->getColSolution(), tCols, colsol);
}
else{
//just take the current solution
lowerObjVal = sSolver->getObjValue();
CoinCopyN(sSolver->getColSolution(), tCols, colsol);
}
delete[] nObjCoeffs;
nObjCoeffs = 0;
delete sSolver;
delete nSolver;
return mcSol(std::make_pair(upperObjVal, lowerObjVal), colsol);
}
else if(beta == 0.0){
/*
fix lower-level objective to current value and
reoptimize wrt to upper-level objective
*/
//OsiSolverInterface * nSolver = new OsiCbcSolverInterface();
OsiSolverInterface * nSolver = new OsiSymSolverInterface();
nSolver->loadProblem(*oSolver->getMatrixByCol(),
oSolver->getColLower(), oSolver->getColUpper(),
oSolver->getObjCoefficients(),
oSolver->getRowLower(), oSolver->getRowUpper());
for(j = 0; j < tCols; j++){
if(oSolver->isInteger(j))
nSolver->setInteger(j);
}
CoinZeroN(nObjCoeffs, tCols);
for(i = 0; i < tCols; i++)
nObjCoeffs[i] = uObjCoeffs[i] * uObjSense;
nSolver->setObjective(nObjCoeffs);
CoinPackedVector objCon;
for(i = 0; i < lCols; i++){
index = lColIndices[i];
objCon.insert(index, lObjCoeffs[i] * lObjSense);
}
nSolver->addRow(objCon, lowerObjVal, lowerObjVal);
if(0){
dynamic_cast<OsiCbcSolverInterface *>
(nSolver)->getModelPtr()->messageHandler()->setLogLevel(0);
}
else{
dynamic_cast<OsiSymSolverInterface *>
(nSolver)->setSymParam("prep_level", -1);
dynamic_cast<OsiSymSolverInterface *>
(nSolver)->setSymParam("verbosity", -2);
dynamic_cast<OsiSymSolverInterface *>
(nSolver)->setSymParam("max_active_nodes", 1);
}
if(0)
nSolver->writeLp("nSolver");
nSolver->branchAndBound();
double * colsol = new double[tCols];
if(nSolver->isProvenOptimal()){
upperObjVal = nSolver->getObjValue();
CoinCopyN(nSolver->getColSolution(), tCols, colsol);
}
else{
upperObjVal = nSolver->getObjValue();
CoinCopyN(nSolver->getColSolution(), tCols, colsol);
}
delete[] nObjCoeffs;
nObjCoeffs = 0;
delete sSolver;
delete nSolver;
return mcSol(std::make_pair(upperObjVal, lowerObjVal), colsol);
}
else{
//no optimality cut needed here. all solutions are supported.
double * colsol = new double[tCols];
CoinCopyN(sSolver->getColSolution(), tCols, colsol);
delete[] nObjCoeffs;
nObjCoeffs = 0;
delete sSolver;
return mcSol(std::make_pair(upperObjVal, lowerObjVal), colsol);
}
}
else{
//FIXME:SHOULD JUST TAKE THIS OUT. DELETE sSolver and remove it from above
nObjCoeffs = 0;
delete[] nObjCoeffs;
delete sSolver;
std::cout << "Subproblem is not proven optimal." << std::endl;
//return NULL;
//abort();
}
}
//#############################################################################
bfSol*
MibSHeuristic::getBilevelSolution1(const double * sol)
{
/*
Find a bilevel feasible solution by solving the LL problem
for a fixed UL solution, given by the UL portion of sol
*/
MibSModel * model = MibSModel_;
OsiSolverInterface * oSolver = model->getSolver();
OsiSolverInterface * lSolver = new OsiCbcSolverInterface(oSolver);
//double uObjSense(model->getSolver()->getObjSense());
double lObjSense(model->getLowerObjSense());
int lCols(model->getLowerDim());
int uCols(model->getUpperDim());
int * lColIndices = model->getLowerColInd();
int uRowNum = model->getUpperRowNum();
int lRowNum = model->getLowerRowNum();
int * uRowIndices = model->getUpperRowInd();
int * lRowIndices = model->getLowerRowInd();
int * uColIndices = model->getUpperColInd();
double * lObjCoeffs = model->getLowerObjCoeffs();
int tCols(uCols + lCols);
int i(0), index(0);
/* delete the UL rows */
lSolver->deleteRows(uRowNum, uRowIndices);
/* Fix the UL variables to their current value in sol */
for(i = 0; i < uCols; i++){
index = uColIndices[i];
lSolver->setColLower(index, sol[index]);
lSolver->setColUpper(index, sol[index]);
}
/* Set the objective to the LL objective coefficients */
double * nObjCoeffs = new double[tCols];
CoinZeroN(nObjCoeffs, tCols);
for(i = 0; i < lCols; i++){
index = lColIndices[i];
nObjCoeffs[index] = lObjCoeffs[i] * lObjSense;
}
lSolver->setObjective(nObjCoeffs);
if(0){
dynamic_cast<OsiCbcSolverInterface *>
(lSolver)->getModelPtr()->messageHandler()->setLogLevel(0);
}
else{
dynamic_cast<OsiSymSolverInterface *>
(lSolver)->setSymParam("prep_level", -1);
dynamic_cast<OsiSymSolverInterface *>
(lSolver)->setSymParam("verbosity", -2);
dynamic_cast<OsiSymSolverInterface *>
(lSolver)->setSymParam("max_active_nodes", 1);
}
lSolver->branchAndBound();
if(lSolver->isProvenOptimal()){
double objVal(0.0);
for(i = 0; i < tCols; i++)
objVal += lSolver->getColSolution()[i] * oSolver->getObjCoefficients()[i];
double * colsol = new double[tCols];
CoinCopyN(lSolver->getColSolution(), tCols, colsol);
bfSol * bfsol =
new bfSol(objVal, colsol);
delete lSolver;
return bfsol;
}
else{
delete lSolver;
return NULL;
}
}
/*
Comments needed
Returns 1 if solution, 0 if not */
int
CbcHeuristicPivotAndFix::solution(double & /*solutionValue*/,
double * /*betterSolution*/)
{
numCouldRun_++; // Todo: Ask JJHF what this for.
std::cout << "Entering Pivot-and-Fix Heuristic" << std::endl;
#ifdef HEURISTIC_INFORM
printf("Entering heuristic %s - nRuns %d numCould %d when %d\n",
heuristicName(),numRuns_,numCouldRun_,when_);
#endif
#ifdef FORNOW
std::cout << "Lucky you! You're in the Pivot-and-Fix Heuristic" << std::endl;
// The struct should be moved to member data
typedef struct {
int numberSolutions;
int maximumSolutions;
int numberColumns;
double ** solution;
int * numberUnsatisfied;
} clpSolution;
double start = CoinCpuTime();
OsiClpSolverInterface * clpSolverOriginal
= dynamic_cast<OsiClpSolverInterface *> (model_->solver());
assert (clpSolverOriginal);
OsiClpSolverInterface *clpSolver(clpSolverOriginal);
ClpSimplex * simplex = clpSolver->getModelPtr();
// Initialize the structure holding the solutions
clpSolution solutions;
// Set typeStruct field of ClpTrustedData struct to one.
// This tells Clp it's "Mahdi!"
ClpTrustedData trustedSolutions;
trustedSolutions.typeStruct = 1;
trustedSolutions.data = &solutions;
solutions.numberSolutions = 0;
solutions.maximumSolutions = 0;
solutions.numberColumns = simplex->numberColumns();
solutions.solution = NULL;
solutions.numberUnsatisfied = NULL;
simplex->setTrustedUserPointer(&trustedSolutions);
// Solve from all slack to get some points
simplex->allSlackBasis();
simplex->primal();
// -------------------------------------------------
// Get the problem information
// - get the number of cols and rows
int numCols = clpSolver->getNumCols();
int numRows = clpSolver->getNumRows();
// - get the right hand side of the rows
const double * rhs = clpSolver->getRightHandSide();
// - find the integer variables
bool * varClassInt = new bool[numCols];
int numInt = 0;
for (int i = 0; i < numCols; i++) {
if (clpSolver->isContinuous(i))
varClassInt[i] = 0;
else {
varClassInt[i] = 1;
numInt++;
}
}
// -Get the rows sense
const char * rowSense;
rowSense = clpSolver->getRowSense();
// -Get the objective coefficients
const double *objCoefficients = clpSolver->getObjCoefficients();
double *originalObjCoeff = new double [numCols];
for (int i = 0; i < numCols; i++)
originalObjCoeff[i] = objCoefficients[i];
// -Get the matrix of the problem
double ** matrix = new double * [numRows];
for (int i = 0; i < numRows; i++) {
matrix[i] = new double[numCols];
for (int j = 0; j < numCols; j++)
matrix[i][j] = 0;
}
const CoinPackedMatrix* matrixByRow = clpSolver->getMatrixByRow();
const double * matrixElements = matrixByRow->getElements();
const int * matrixIndices = matrixByRow->getIndices();
const int * matrixStarts = matrixByRow->getVectorStarts();
for (int j = 0; j < numRows; j++) {
for (int i = matrixStarts[j]; i < matrixStarts[j+1]; i++) {
matrix[j][matrixIndices[i]] = matrixElements[i];
}
}
// The newObj is the randomly perturbed constraint used to find new
// corner points
double * newObj = new double [numCols];
// Set the random seed
srand ( time(NULL) + 1);
int randNum;
// We're going to add a new row to the LP formulation
// after finding each new solution.
// Adding a new row requires the new elements and the new indices.
// The elements are original objective function coefficients.
// The indicies are the (dense) columns indices stored in addRowIndex.
// The rhs is the value of the new solution stored in solutionValue.
int * addRowIndex = new int[numCols];
for (int i = 0; i < numCols; i++)
addRowIndex[i] = i;
// The number of feasible solutions found by the PF heuristic.
// This controls the return code of the solution() method.
int numFeasibles = 0;
// Shuffle the rows
int * index = new int [numRows];
for (int i = 0; i < numRows; i++)
index[i] = i;
for (int i = 0; i < numRows; i++) {
int temp = index[i];
int randNumTemp = i + (rand() % (numRows - i));
index[i] = index[randNumTemp];
index[randNumTemp] = temp;
}
// In the clpSolution struct, we store a lot of column solutions.
// For each perturb objective, we store the solution from each
// iteration of the LP solve.
// For each perturb objective, we look at the collection of
// solutions to do something extremly intelligent :-)
// We could (and should..and will :-) wipe out the block of
// solutions when we're done with them. But for now, we just move on
// and store the next block of solutions for the next (perturbed)
// objective.
// The variable startIndex tells us where the new block begins.
int startIndex = 0;
// At most "fixThreshold" number of integer variables can be unsatisfied
// for calling smallBranchAndBound().
// The PF Heuristic only fixes fixThreshold number of variables to
// their integer values. Not more. Not less. The reason is to give
// the smallBB some opportunity to find better solutions. If we fix
// everything it might be too many (leading the heuristic to come up
// with infeasibility rather than a useful result).
// (This is an important paramater. And it is dynamically set.)
double fixThreshold;
/*
if(numInt > 400)
fixThreshold = 17*sqrt(numInt);
if(numInt<=400 && numInt>100)
fixThreshold = 5*sqrt(numInt);
if(numInt<=100)
fixThreshold = 4*sqrt(numInt);
*/
// Initialize fixThreshold based on the number of integer
// variables
if (numInt <= 100)
fixThreshold = .35 * numInt;
if (numInt > 100 && numInt < 1000)
fixThreshold = .85 * numInt;
if (numInt >= 1000)
fixThreshold = .1 * numInt;
// Whenever the dynamic system for changing fixThreshold
// kicks in, it changes the parameter by the
// fixThresholdChange amount.
// (The 25% should be member data and tuned. Another paper!)
double fixThresholdChange = 0.25 * fixThreshold;
// maxNode is the maximum number of nodes we allow smallBB to
// search. It's initialized to 400 and changed dynamically.
// The 400 should be member data, if we become virtuous.
int maxNode = 400;
// We control the decision to change maxNode through the boolean
// variable changeMaxNode. The boolean variable is initialized to
// true and gets set to false under a condition (and is never true
// again.)
// It's flipped off and stays off (in the current incarnation of PF)
bool changeMaxNode = 1;
// The sumReturnCode is used for the dynamic system that sets
// fixThreshold and changeMaxNode.
//
// We track what's happening in sumReturnCode. There are 8 switches.
// The first 5 switches corresponds to a return code for smallBB.
//
// We want to know how many times we consecutively get the same
// return code.
//
// If "good" return codes are happening often enough, we're happy.
//
// If a "bad" returncodes happen consecutively, we want to
// change something.
//
// The switch 5 is the number of times PF didn't call smallBB
// becuase the number of integer variables that took integer values
// was less than fixThreshold.
//
// The swicth 6 was added for a brilliant idea...to be announced
// later (another paper!)
//
// The switch 7 is the one that changes the max node. Read the
// code. (Todo: Verbalize the brilliant idea for the masses.)
//
int sumReturnCode[8];
/*
sumReturnCode[0] ~ -1 --> problem too big for smallBB
sumReturnCode[1] ~ 0 --> smallBB not finshed and no soln
sumReturnCode[2] ~ 1 --> smallBB not finshed and there is a soln
sumReturnCode[3] ~ 2 --> smallBB finished and no soln
sumReturnCode[4] ~ 3 --> smallBB finished and there is a soln
sumReturnCode[5] ~ didn't call smallBranchAndBound too few to fix
sumReturnCode[6] ~ didn't call smallBranchAndBound too many unsatisfied
sumReturnCode[7] ~ the same as sumReturnCode[1] but becomes zero just if the returnCode is not 0
*/
for (int i = 0; i < 8; i++)
sumReturnCode[i] = 0;
int * colIndex = new int[numCols];
for (int i = 0; i < numCols; i++)
colIndex[i] = i;
double cutoff = COIN_DBL_MAX;
bool didMiniBB;
// Main loop
for (int i = 0; i < numRows; i++) {
// track the number of mini-bb for the dynamic threshold setting
didMiniBB = 0;
for (int k = startIndex; k < solutions.numberSolutions; k++)
//if the point has 0 unsatisfied variables; make sure it is
//feasible. Check integer feasiblity and constraints.
if (solutions.numberUnsatisfied[k] == 0) {
double feasibility = 1;
//check integer feasibility
for (int icol = 0; icol < numCols; icol++) {
double closest = floor(solutions.solution[k][icol] + 0.5);
if (varClassInt[icol] && (fabs(solutions.solution[k][icol] - closest) > 1e-6)) {
feasibility = 0;
break;
}
}
//check if the solution satisfies the constraints
for (int irow = 0; irow < numRows; irow++) {
double lhs = 0;
for (int j = 0; j < numCols; j++)
lhs += matrix[irow][j] * solutions.solution[k][j];
if (rowSense[irow] == 'L' && lhs > rhs[irow] + 1e-6) {
feasibility = 0;
break;
}
if (rowSense[irow] == 'G' && lhs < rhs[irow] - 1e-6) {
feasibility = 0;
break;
}
if (rowSense[irow] == 'E' && (lhs - rhs[irow] > 1e-6 || lhs - rhs[irow] < -1e-6)) {
feasibility = 0;
break;
}
}
//if feasible, find the objective value and set the cutoff
// for the smallBB and add a new constraint to the LP
// (and update the best solution found so far for the
// return arguments)
if (feasibility) {
double objectiveValue = 0;
for (int j = 0; j < numCols; j++)
objectiveValue += solutions.solution[k][j] * originalObjCoeff[j];
cutoff = objectiveValue;
clpSolver->addRow(numCols, addRowIndex, originalObjCoeff, -COIN_DBL_MAX, cutoff);
// Todo: pick up the best solution in the block (not
// the last).
solutionValue = objectiveValue;
for (int m = 0; m < numCols; m++)
betterSolution[m] = solutions.solution[k][m];
numFeasibles++;
}
}
// Go through the block of solution and decide if to call smallBB
for (int k = startIndex; k < solutions.numberSolutions; k++) {
if (solutions.numberUnsatisfied[k] <= fixThreshold) {
// get new copy
OsiSolverInterface * newSolver;
newSolver = new OsiClpSolverInterface(*clpSolver);
newSolver->setObjSense(1);
newSolver->setObjective(originalObjCoeff);
int numberColumns = newSolver->getNumCols();
int numFixed = 0;
// Fix the first fixThreshold number of integer vars
// that are satisfied
for (int iColumn = 0 ; iColumn < numberColumns ; iColumn++) {
if (newSolver->isInteger(iColumn)) {
double value = solutions.solution[k][iColumn];
double intValue = floor(value + 0.5);
if (fabs(value - intValue) < 1.0e-5) {
newSolver->setColLower(iColumn, intValue);
newSolver->setColUpper(iColumn, intValue);
numFixed++;
if (numFixed > numInt - fixThreshold)
break;
}
}
}
COIN_DETAIL_PRINT(printf("numFixed: %d\n", numFixed));
COIN_DETAIL_PRINT(printf("fixThreshold: %f\n", fixThreshold));
COIN_DETAIL_PRINT(printf("numInt: %d\n", numInt));
double *newSolution = new double[numCols];
double newSolutionValue;
// Call smallBB on the modified problem
int returnCode = smallBranchAndBound(newSolver, maxNode, newSolution,
newSolutionValue, cutoff, "mini");
// If smallBB found a solution, update the better
// solution and solutionValue (we gave smallBB our
// cutoff, so it only finds improving solutions)
if (returnCode == 1 || returnCode == 3) {
numFeasibles ++;
solutionValue = newSolutionValue;
for (int m = 0; m < numCols; m++)
betterSolution[m] = newSolution[m];
COIN_DETAIL_PRINT(printf("cutoff: %f\n", newSolutionValue));
COIN_DETAIL_PRINT(printf("time: %.2lf\n", CoinCpuTime() - start));
}
didMiniBB = 1;
COIN_DETAIL_PRINT(printf("returnCode: %d\n", returnCode));
//Update sumReturnCode array
for (int iRC = 0; iRC < 6; iRC++) {
if (iRC == returnCode + 1)
sumReturnCode[iRC]++;
else
sumReturnCode[iRC] = 0;
}
if (returnCode != 0)
sumReturnCode[7] = 0;
else
sumReturnCode[7]++;
if (returnCode == 1 || returnCode == 3) {
cutoff = newSolutionValue;
clpSolver->addRow(numCols, addRowIndex, originalObjCoeff, -COIN_DBL_MAX, cutoff);
COIN_DETAIL_PRINT(printf("******************\n\n*****************\n"));
}
break;
}
}
if (!didMiniBB && solutions.numberSolutions - startIndex > 0) {
sumReturnCode[5]++;
for (int iRC = 0; iRC < 5; iRC++)
sumReturnCode[iRC] = 0;
}
//Change "fixThreshold" if needed
// using the data we've recorded in sumReturnCode
if (sumReturnCode[1] >= 3)
fixThreshold -= fixThresholdChange;
if (sumReturnCode[7] >= 3 && changeMaxNode) {
maxNode *= 5;
changeMaxNode = 0;
}
if (sumReturnCode[3] >= 3 && fixThreshold < 0.95 * numInt)
fixThreshold += fixThresholdChange;
if (sumReturnCode[5] >= 4)
fixThreshold += fixThresholdChange;
if (sumReturnCode[0] > 3)
fixThreshold -= fixThresholdChange;
startIndex = solutions.numberSolutions;
//Check if the maximum iterations limit is reached
// rlh: Ask John how this is working with the change to trustedUserPtr.
if (solutions.numberSolutions > 20000)
break;
// The first time in this loop PF solves orig LP.
//Generate the random objective function
randNum = rand() % 10 + 1;
randNum = fmod(randNum, 2);
for (int j = 0; j < numCols; j++) {
if (randNum == 1)
if (fabs(matrix[index[i]][j]) < 1e-6)
newObj[j] = 0.1;
else
newObj[j] = matrix[index[i]][j] * 1.1;
else if (fabs(matrix[index[i]][j]) < 1e-6)
newObj[j] = -0.1;
else
newObj[j] = matrix[index[i]][j] * 0.9;
}
clpSolver->setObjective(newObj);
if (rowSense[i] == 'L')
clpSolver->setObjSense(-1);
else
// Todo #1: We don't need to solve the LPs to optimality.
// We just need corner points.
// There's a problem in stopping Clp that needs to be looked
// into. So for now, we solve optimality.
clpSolver->setObjSense(1);
// simplex->setMaximumIterations(100);
clpSolver->getModelPtr()->primal(1);
// simplex->setMaximumIterations(100000);
#ifdef COIN_DETAIL
printf("cutoff: %f\n", cutoff);
printf("time: %.2f\n", CoinCpuTime() - start);
for (int iRC = 0; iRC < 8; iRC++)
printf("%d ", sumReturnCode[iRC]);
printf("\nfixThreshold: %f\n", fixThreshold);
printf("numInt: %d\n", numInt);
printf("\n---------------------------------------------------------------- %d\n", i);
#endif
//temp:
if (i > 3) break;
}
COIN_DETAIL_PRINT(printf("Best Feasible Found: %f\n", cutoff));
COIN_DETAIL_PRINT(printf("Total time: %.2f\n", CoinCpuTime() - start));
if (numFeasibles == 0) {
return 0;
}
// We found something better
std::cout << "See you soon! You're leaving the Pivot-and-Fix Heuristic" << std::endl;
std::cout << std::endl;
return 1;
#endif
return 0;
}
void
HeuristicInnerApproximation::extractInnerApproximation(Bonmin::OsiTMINLPInterface & nlp, OsiSolverInterface &si,
const double * x, bool getObj) {
printf("************ Start extracting inner approx");
int n;
int m;
int nnz_jac_g;
int nnz_h_lag;
Ipopt::TNLP::IndexStyleEnum index_style;
Bonmin::TMINLP2TNLP * problem = nlp.problem();
//Get problem information
problem->get_nlp_info(n, m, nnz_jac_g, nnz_h_lag, index_style);
Bonmin::vector<int> jRow(nnz_jac_g);
Bonmin::vector<int> jCol(nnz_jac_g);
Bonmin::vector<double> jValues(nnz_jac_g);
problem->eval_jac_g(n, NULL, 0, m, nnz_jac_g, jRow(), jCol(), NULL);
if(index_style == Ipopt::TNLP::FORTRAN_STYLE)//put C-style
{
for(int i = 0 ; i < nnz_jac_g ; i++){
jRow[i]--;
jCol[i]--;
}
}
//get Jacobian
problem->eval_jac_g(n, x, 1, m, nnz_jac_g, NULL, NULL,
jValues());
Bonmin::vector<double> g(m);
problem->eval_g(n, x, 1, m, g());
Bonmin::vector<int> nonLinear(m);
//store non linear constraints (which are to be removed from IA)
int numNonLinear = 0;
const double * rowLower = nlp.getRowLower();
const double * rowUpper = nlp.getRowUpper();
const double * colLower = nlp.getColLower();
const double * colUpper = nlp.getColUpper();
assert(m == nlp.getNumRows());
double infty = si.getInfinity();
double nlp_infty = nlp.getInfinity();
Bonmin::vector<Ipopt::TNLP::LinearityType> constTypes(m);
Bonmin::vector<Ipopt::TNLP::LinearityType> varTypes(n);
problem->get_constraints_linearity(m, constTypes());
problem->get_variables_linearity(n, varTypes());
for (int i = 0; i < m; i++) {
if (constTypes[i] == Ipopt::TNLP::NON_LINEAR) {
nonLinear[numNonLinear++] = i;
}
}
Bonmin::vector<double> rowLow(m - numNonLinear);
Bonmin::vector<double> rowUp(m - numNonLinear);
int ind = 0;
for (int i = 0; i < m; i++) {
if (constTypes[i] != Ipopt::TNLP::NON_LINEAR) {
if (rowLower[i] > -nlp_infty) {
// printf("Lower %g ", rowLower[i]);
rowLow[ind] = (rowLower[i]);
} else
rowLow[ind] = -infty;
if (rowUpper[i] < nlp_infty) {
// printf("Upper %g ", rowUpper[i]);
rowUp[ind] = (rowUpper[i]);
} else
rowUp[ind] = infty;
ind++;
}
}
CoinPackedMatrix mat(true, jRow(), jCol(), jValues(), nnz_jac_g);
mat.setDimensions(m, n); // In case matrix was empty, this should be enough
//remove non-linear constraints
mat.deleteRows(numNonLinear, nonLinear());
int numcols = nlp.getNumCols();
Bonmin::vector<double> obj(numcols);
for (int i = 0; i < numcols; i++)
obj[i] = 0.;
si.loadProblem(mat, nlp.getColLower(), nlp.getColUpper(),
obj(), rowLow(), rowUp());
const Bonmin::TMINLP::VariableType* variableType = problem->var_types();
for (int i = 0; i < n; i++) {
if ((variableType[i] == Bonmin::TMINLP::BINARY) || (variableType[i] == Bonmin::TMINLP::INTEGER))
si.setInteger(i);
}
if (getObj) {
bool addObjVar = false;
if (problem->hasLinearObjective()) {
double zero;
Bonmin::vector<double> x0(n, 0.);
problem->eval_f(n, x0(), 1, zero);
si.setDblParam(OsiObjOffset, -zero);
//Copy the linear objective and don't create a dummy variable.
problem->eval_grad_f(n, x, 1, obj());
si.setObjective(obj());
} else {
addObjVar = true;
}
if (addObjVar) {
nlp.addObjectiveFunction(si, x);
}
}
// Hassan IA initial description
int InnerDesc = 1;
if (InnerDesc == 1) {
OsiCuts cs;
double * p = CoinCopyOfArray(colLower, n);
double * pp = CoinCopyOfArray(colLower, n);
double * up = CoinCopyOfArray(colUpper, n);
for (int i = 0; i < n; i++) {
if (p[i] < -1e3){
p[i] = pp[i] = -1e3;
}
if (up[i] > 1e2){
up[i] = 1e2;
}
}
const int& nbAp = nbAp_;
printf("Generating approximation with %i points.\n", nbAp);
std::vector<double> step(n);
int n_lin = 0;
for (int i = 0; i < n; i++) {
//if ((variableType[i] == Bonmin::TMINLP::BINARY) || (variableType[i] == Bonmin::TMINLP::INTEGER)) {
if (varTypes[i] == Ipopt::TNLP::LINEAR) {
n_lin ++;
step[i] = 0;
p[i] = pp[i] = up[i] = 0;
}
else {
step[i] = (up[i] - p[i]) / (nbAp);
}
}
printf("Number of linears %i\n", n_lin);
for (int j = 1; j < nbAp; j++) {
for (int i = 0; i < n; i++) {
pp[i] += step[i];
}
for (int i = 0; (i < m ); i++) {
if (constTypes[i] == Ipopt::TNLP::LINEAR) continue;
bool status = getMyInnerApproximation(nlp, cs, i, p, pp);// Generate a chord connecting the two points
if(status == false){
printf("Error in generating inner approximation\n");
exit(1);
}
}
std::copy(pp, pp+n, p);
}
for(int i = 0; (i< m); i++) {
if (constTypes[i] == Ipopt::TNLP::LINEAR) continue;
getMyInnerApproximation(nlp, cs, i, p, up);// Generate a chord connecting the two points
}
delete [] p;
delete [] pp;
delete [] up;
si.applyCuts(cs);
}
printf("************ Done extracting inner approx ********");
}
void
HeuristicInnerApproximation::extractInnerApproximation(OsiTMINLPInterface & nlp, OsiSolverInterface &si,
const double * x, bool getObj) {
int n;
int m;
int nnz_jac_g;
int nnz_h_lag;
Ipopt::TNLP::IndexStyleEnum index_style;
TMINLP2TNLP * problem = nlp.problem();
//Get problem information
problem->get_nlp_info(n, m, nnz_jac_g, nnz_h_lag, index_style);
vector<int> jRow(nnz_jac_g);
vector<int> jCol(nnz_jac_g);
vector<double> jValues(nnz_jac_g);
problem->eval_jac_g(n, NULL, 0, m, nnz_jac_g, jRow(), jCol(), NULL);
if(index_style == Ipopt::TNLP::FORTRAN_STYLE)//put C-style
{
for(int i = 0 ; i < nnz_jac_g ; i++){
jRow[i]--;
jCol[i]--;
}
}
//get Jacobian
problem->eval_jac_g(n, x, 1, m, nnz_jac_g, NULL, NULL,
jValues());
vector<double> g(m);
problem->eval_g(n, x, 1, m, g());
vector<int> nonLinear(m);
//store non linear constraints (which are to be removed from IA)
int numNonLinear = 0;
const double * rowLower = nlp.getRowLower();
const double * rowUpper = nlp.getRowUpper();
const double * colLower = nlp.getColLower();
const double * colUpper = nlp.getColUpper();
assert(m == nlp.getNumRows());
double infty = si.getInfinity();
double nlp_infty = nlp.getInfinity();
vector<Ipopt::TNLP::LinearityType> constTypes(m);
problem->get_constraints_linearity(m, constTypes());
for (int i = 0; i < m; i++) {
if (constTypes[i] == Ipopt::TNLP::NON_LINEAR) {
nonLinear[numNonLinear++] = i;
}
}
vector<double> rowLow(m - numNonLinear);
vector<double> rowUp(m - numNonLinear);
int ind = 0;
for (int i = 0; i < m; i++) {
if (constTypes[i] != Ipopt::TNLP::NON_LINEAR) {
if (rowLower[i] > -nlp_infty) {
// printf("Lower %g ", rowLower[i]);
rowLow[ind] = (rowLower[i]);
} else
rowLow[ind] = -infty;
if (rowUpper[i] < nlp_infty) {
// printf("Upper %g ", rowUpper[i]);
rowUp[ind] = (rowUpper[i]);
} else
rowUp[ind] = infty;
ind++;
}
}
CoinPackedMatrix mat(true, jRow(), jCol(), jValues(), nnz_jac_g);
mat.setDimensions(m, n); // In case matrix was empty, this should be enough
//remove non-linear constraints
mat.deleteRows(numNonLinear, nonLinear());
int numcols = nlp.getNumCols();
vector<double> obj(numcols);
for (int i = 0; i < numcols; i++)
obj[i] = 0.;
si.loadProblem(mat, nlp.getColLower(), nlp.getColUpper(),
obj(), rowLow(), rowUp());
const Bonmin::TMINLP::VariableType* variableType = problem->var_types();
for (int i = 0; i < n; i++) {
if ((variableType[i] == TMINLP::BINARY) || (variableType[i]
== TMINLP::INTEGER))
si.setInteger(i);
}
if (getObj) {
bool addObjVar = false;
if (problem->hasLinearObjective()) {
double zero;
vector<double> x0(n, 0.);
problem->eval_f(n, x0(), 1, zero);
si.setDblParam(OsiObjOffset, -zero);
//Copy the linear objective and don't create a dummy variable.
problem->eval_grad_f(n, x, 1, obj());
si.setObjective(obj());
} else {
addObjVar = true;
}
if (addObjVar) {
nlp.addObjectiveFunction(si, x);
}
}
// Hassan IA initial description
int InnerDesc = 1;
if (InnerDesc == 1) {
OsiCuts cs;
double * p = CoinCopyOfArray(colLower, n);
double * pp = CoinCopyOfArray(colLower, n);
double * up = CoinCopyOfArray(colUpper, n);
const int& nbAp = nbAp_;
std::vector<int> nbG(m, 0);// Number of generated points for each nonlinear constraint
std::vector<double> step(n);
for (int i = 0; i < n; i++) {
if (colUpper[i] > 1e08) {
up[i] = 0;
}
if (colUpper[i] > 1e08 || colLower[i] < -1e08 || (variableType[i]
== TMINLP::BINARY) || (variableType[i] == TMINLP::INTEGER)) {
step[i] = 0;
} else
step[i] = (up[i] - colLower[i]) / (nbAp);
if (colLower[i] < -1e08) {
p[i] = 0;
pp[i] = 0;
}
}
vector<double> g_p(m);
vector<double> g_pp(m);
for (int j = 1; j <= nbAp; j++) {
for (int i = 0; i < n; i++) {
pp[i] += step[i];
}
problem->eval_g(n, p, 1, m, g_p());
problem->eval_g(n, pp, 1, m, g_pp());
double diff = 0;
int varInd = 0;
for (int i = 0; (i < m && constTypes[i] == Ipopt::TNLP::NON_LINEAR); i++) {
if (varInd == n - 1)
varInd = 0;
diff = std::abs(g_p[i] - g_pp[i]);
if (nbG[i] < nbAp - 1) {
getMyInnerApproximation(nlp, cs, i, p, pp);// Generate a chord connecting the two points
p[varInd] = pp[varInd];
nbG[i]++;
}
varInd++;
}
}
for(int i = 0; (i< m && constTypes[i] == Ipopt::TNLP::NON_LINEAR); i++) {
// getConstraintOuterApproximation(cs, i, colUpper, NULL, true);// Generate Tangents at current point
getMyInnerApproximation(nlp, cs, i, p, up);// Generate a chord connecting the two points
}
delete [] p;
delete [] pp;
delete [] up;
si.applyCuts(cs);
}
}
