org.ojalgo.optimisation.integer.NodeSolver Maven / Gradle / Ivy
Go to download
Show more of this group Show more artifacts with this name
Show all versions of ojalgo Show documentation
Show all versions of ojalgo Show documentation
oj! Algorithms - ojAlgo - is Open Source Java code that has to do with mathematics, linear algebra and optimisation.
/*
* Copyright 1997-2022 Optimatika
*
* Permission is hereby granted, free of charge, to any person obtaining a copy
* of this software and associated documentation files (the "Software"), to deal
* in the Software without restriction, including without limitation the rights
* to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
* copies of the Software, and to permit persons to whom the Software is
* furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included in
* all copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
* AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
* SOFTWARE.
*/
package org.ojalgo.optimisation.integer;
import java.math.BigDecimal;
import java.math.RoundingMode;
import java.util.Collection;
import java.util.Iterator;
import java.util.Map.Entry;
import java.util.concurrent.atomic.AtomicInteger;
import org.ojalgo.equation.Equation;
import org.ojalgo.function.constant.BigMath;
import org.ojalgo.function.special.MissingMath;
import org.ojalgo.netio.BasicLogger;
import org.ojalgo.optimisation.Expression;
import org.ojalgo.optimisation.ExpressionsBasedModel;
import org.ojalgo.optimisation.IntermediateSolver;
import org.ojalgo.optimisation.ModelEntity;
import org.ojalgo.optimisation.UpdatableSolver;
import org.ojalgo.structure.Structure1D.IntIndex;
import org.ojalgo.type.context.NumberContext;
import org.ojalgo.type.keyvalue.EntryPair;
public final class NodeSolver extends IntermediateSolver {
private static final NumberContext PRECISION = NumberContext.of(12);
private static final NumberContext COEFFICIENT = PRECISION.withMode(RoundingMode.CEILING);
private static final AtomicInteger COUNTER = new AtomicInteger();
private static final boolean DEBUG = false;
private static final NumberContext DYNANISM = NumberContext.of(8);
private static final NumberContext LIMIT = PRECISION.withMode(RoundingMode.FLOOR);
private static final NumberContext SCALE = NumberContext.of(14);
NodeSolver(final ExpressionsBasedModel model) {
super(model);
}
boolean generateCuts(final ModelStrategy strategy) {
boolean retVal = this.generateCuts(strategy, this.getModel());
if (retVal) {
this.reset();
}
return retVal;
}
boolean generateCuts(final ModelStrategy strategy, final ExpressionsBasedModel target) {
if (!this.isSolved()) {
return false;
}
ExpressionsBasedModel model = this.getModel();
Result result = this.getResult();
long nbConstr = model.constraints().count();
if (this.getSolver() instanceof UpdatableSolver) {
UpdatableSolver solver = (UpdatableSolver) this.getSolver();
UpdatableSolver.EntityMap entityMap = solver.getEntityMap();
if (entityMap != null) {
int nbProblVars = entityMap.countVariables();
int nbProblInts = strategy.countIntegerVariables();
int nbSlackVars = entityMap.countSlackVariables();
boolean[] integers = new boolean[nbProblVars + nbSlackVars];
for (int i = 0; i < nbProblInts; i++) {
int indexInModel = strategy.getIndex(i);
int indexInSolver = this.getIndexInSolver(indexInModel);
if (indexInSolver >= 0) {
integers[indexInSolver] = true;
}
}
for (int i = 0; i < nbSlackVars; i++) {
EntryPair, ConstraintType> slack = entityMap.getSlack(i);
ModelEntity key = slack.getKey();
boolean integer = key.isInteger();
integers[nbProblVars + i] = integer;
}
Collection potentialCuts = solver.generateCutCandidates(strategy.getGMICutConfiguration().fractionality, integers);
for (Equation equation : potentialCuts) {
String name = "CUT_GMI_" + equation.index + "_" + COUNTER.incrementAndGet();
if (DEBUG) {
BasicLogger.debug("{} {}", name, equation.toString());
}
Expression cut = target.addExpression(name);
cut.lower(BigMath.ONE);
for (int j = 0; j < nbProblVars; j++) {
int mj = entityMap.indexOf(j);
double aj = equation.doubleValue(j);
if (!SCALE.isZero(aj)) {
if (entityMap.isNegated(j)) {
cut.add(mj, -aj);
} else {
cut.add(mj, aj);
}
}
}
for (int j = 0; j < entityMap.countSlackVariables(); j++) {
double aj = equation.doubleValue(nbProblVars + j);
if (!SCALE.isZero(aj)) {
EntryPair, ConstraintType> pair = entityMap.getSlack(j);
ModelEntity entity = pair.getKey();
ConstraintType type = pair.getValue();
BigDecimal coefficient = BigDecimal.valueOf(aj);
BigDecimal adjusted = entity.adjust(coefficient);
if (ConstraintType.LOWER.equals(type)) {
BigDecimal factor = adjusted;
BigDecimal limit = entity.getLowerLimit();
BigDecimal shift = limit.multiply(factor);
cut.shift(shift);
entity.addTo(cut, factor);
}
if (ConstraintType.UPPER.equals(type)) {
BigDecimal factor = adjusted.negate();
BigDecimal limit = entity.getUpperLimit();
BigDecimal shift = limit.multiply(factor);
cut.shift(shift);
entity.addTo(cut, factor);
}
}
}
BigDecimal cRHS = cut.getLowerLimit();
// The cut violation is always 1.0
// The relative violation is 1.0 relative to the RHS
// We only need to check that the RHS is not too large
// The violation configuration property is the largest allowed RHS
BigDecimal violation = strategy.getGMICutConfiguration().violation;
if (cRHS.abs().compareTo(violation) > 0) {
target.removeExpression(name);
if (DEBUG) {
BasicLogger.debug("\tViolation small! {}", cRHS);
}
continue;
}
BigDecimal cLargest = BigMath.ONE;
for (Entry entry : cut.getLinearEntrySet()) {
cLargest = cLargest.max(entry.getValue().abs());
}
BigDecimal cSmallest = BigMath.VERY_POSITIVE;
for (Iterator> iterator = cut.getLinearEntrySet().iterator(); iterator.hasNext();) {
Entry entry = iterator.next();
BigDecimal cValue = entry.getValue();
if (!PRECISION.isSmall(cLargest, cValue)) {
cSmallest = cSmallest.min(cValue.abs());
entry.setValue(COEFFICIENT.enforce(cValue));
} else {
cRHS = cRHS.subtract(cValue.multiply(result.get(entry.getKey().index)));
iterator.remove();
}
}
cRHS = LIMIT.enforce(cRHS);
cut.lower(cRHS);
if (DEBUG) {
BigDecimal cRatio = MissingMath.divide(cLargest, cSmallest);
BigDecimal cEvaluated = cut.evaluate(result);
BasicLogger.debug("\tLargest={}, Smallest={}, Ratio={}: {} < {}", cLargest, cSmallest, cRatio, cRHS, cEvaluated);
}
if (DYNANISM.isSmall(cLargest, cSmallest)) {
target.removeExpression(name);
if (DEBUG) {
BasicLogger.debug("\tDynanism large! {} >> {}", cLargest, cSmallest);
}
} else if (model.checkSimilarity(cut)) {
target.removeExpression(name);
if (DEBUG) {
BasicLogger.debug("\tCut similar to current constraint!");
}
} else {
// Accept this cut!
cut.tighten();
if (DEBUG) {
BasicLogger.debug("\t{}", cut);
}
}
if (DEBUG) {
BasicLogger.debug("\t{} < {}", cut.getLowerLimit(), cut.getLinearEntrySet());
}
if (target.options.logger_detailed && target.options.logger_appender != null) {
target.options.logger_appender.println("{}: {} < {}", name, cut.getLowerLimit(), cut.getLinearEntrySet());
}
}
}
}
return nbConstr != model.constraints().count();
}
}