private void RunEpsilonMethod(int minSoftConsViol, int maxSoftConsViol, double minCost, double maxCost)
{
List <MultiResult> pareto;
var stepsize = (double)1 / (Steps); //(double)1 / (Steps + 1);
var proxobj = 0;
if (EpsilonOnQuality)
{
_model.SetObjective(0, 1, proxobj);
}
else
{
_model.SetObjective(1, 0, proxobj);
}
if (LocalBranching > 0)
{
_model.SetProxConstraint(LocalBranching);
}
for (var alpha = EpsilonRelaxing ? stepsize : 1 - stepsize;
alpha >= 0 && alpha <= 1;
//maybe we maybe should solve the last one again.. this seems to give improvementts.
alpha += (EpsilonRelaxing ? 1 : -1) * stepsize)
{
Console.WriteLine($"Elapsed time: {_stopwatch.Elapsed} -- {_stopwatch.ElapsedMilliseconds}ms");
OptimizeSubProblem(minSoftConsViol, maxSoftConsViol, minCost, maxCost, alpha);
}
if (DoubleSweep)
{
pareto = GetParetoPoints(_multiResults);
var beforepoints = pareto.Count;
DisplaySolutions(pareto);
//Check if new corner points
minCost = pareto.Min(r => r.Cost);
maxCost = pareto.Max(r => r.Cost);
minSoftConsViol = pareto.Min(r => r.SoftObjective);
maxSoftConsViol = pareto.Max(r => r.SoftObjective);
Console.WriteLine("Running opposite direction");
if (LocalBranching > 0)
{
_model.SetProxConstraint(LocalBranching);
}
for (var alpha = !EpsilonRelaxing ? stepsize : 1 - stepsize;
alpha > 0 && alpha < 1;
//maybe we maybe should solve the last one again.. this seems to give improvementts.
alpha += (!EpsilonRelaxing ? 1 : -1) * stepsize)
{
Console.WriteLine($"Elapsed time: {_stopwatch.Elapsed} -- {_stopwatch.ElapsedMilliseconds}ms");
OptimizeSubProblem(minSoftConsViol, maxSoftConsViol, minCost, maxCost, alpha);
}
pareto = GetParetoPoints(_multiResults);
Console.WriteLine($"Done with oposite. New points added: {pareto.Count - beforepoints}");
}
}
public void TestLocalBranchingConstraint(bool usecurr, int N)
{
var data = Data.ReadXml(dataPath + ITC_Comp05, "120", "4");
//var formulation = ProblemFormulation.UD2;
//Formulation.AvailabilityHardConstraint = false;
//Formulation.OverbookingAllowed = 0.35;
var modelParameters = new CCTModel.MIPModelParameters()
{
UseStageIandII = false
};
var model = new CCTModel(data, formulation, modelParameters);
model.Optimize(30 * 3, 0.99);
model.SetProxConstraint(N);
var watch = Stopwatch.StartNew();
while (watch.Elapsed.TotalSeconds < TimelimitStageI)
{
model.SetProximityOrigin(usecurr);
model.Optimize(15);
}
}
public List <Tuple <int, Solution, int> > Run()
{
var before = (int)solutionBefore.Objective;
model.SetObjective(0, 0, 1);
//model.Fixsol(false);
//model.SetProxConstraint(10);
var feasible = model.Optimize(Timelimit);
if (!feasible)
{
return(new List <Tuple <int, Solution, int> >());
}
var minperb = (int)model.Objective;
//Find best objective
model.SetObjective(1, 0, 0);
var sol = new List <Tuple <int, Solution, int> >();
var currentObjective = 0;
for (var pertubations = minperb; pertubations <= minperb + ExtraPerubations && pertubations <= MaxTotalPertubations; pertubations++)
{
Console.WriteLine($"pertubations = {pertubations}");
model.SetProxConstraint(pertubations);
if (PerturbationObjectEqual)
{
model.SetProcConstraintSenseEqual(true);
}
// if (RemoveCurrentSolutionAfteritt) model.b
//model.Optimize();
int sols;
for (sols = 0; sols < SolPerPertubation; sols++)
{
if (!model.Optimize(Timelimit))
{
if (AbortSolverWhenNoImprovement)
{
break; //infeasibles
}
else
{
continue;
}
}
Console.WriteLine("new solution");
//constraint on objective function
// if (model.ObjSoftCons == prev) break;
currentObjective = model.ObjSoftCons;
model.SetQualConstraint(currentObjective);
var solututionAfter = new Solution(data, formulation);
solututionAfter.SetAssignments(model.GetAssignments());
Console.WriteLine($"Before: {before}\n" +
$"After: {model.ObjSoftCons}\n" +
$"Perbs: {pertubations}");
Console.WriteLine(solutionBefore.AnalyzeSolution());
Console.WriteLine(solututionAfter.AnalyzeSolution());
var solution = new Solution(data, formulation);
solution.SetAssignments(model.GetAssignments());
solution.AnalyzeSolution();
sol.Add(Tuple.Create(pertubations, solution, model.RunTime));
DisplayPertubations(solutionBefore._assignments.ToList(), solututionAfter._assignments.ToList());
//to pertubate more.
model.RemoveCurrentSolution();
}
if (sols == 0)
{
break; //previous pertubations didnt find anything
}
model.SetQualConstraint(currentObjective - 1);
}
return(sol);
}
public List <Tuple <int, Solution, int, int> > Run()
{
var starttime = DateTime.Now;
var before = (int)solutionBefore.Objective;
model.SetObjective(0, 0, 1);
//model.Fixsol(false);
//model.SetProxConstraint(10);
var feasible = model.Optimize(Timelimit);
if (!feasible)
{
return(new List <Tuple <int, Solution, int, int> >());
}
var minperb = (int)model.Objective;
//Find best objective
model.SetObjective(1, 0, 0);
var sol = new List <Tuple <int, Solution, int, int> >();
var currentObjective = 0;
var maxPerturbations = Math.Max(minperb + ExtraPerubations, MaxTotalPertubations);
for (var pertubations = minperb; pertubations <= maxPerturbations; pertubations++)
{
Console.WriteLine($"pertubations = {pertubations}");
model.SetProxConstraint(pertubations);
if (PerturbationObjectEqual)
{
model.SetProcConstraintSenseEqual(true);
}
// if (RemoveCurrentSolutionAfteritt) model.b
//model.Optimize();
int sols;
for (sols = 0; sols < SolPerPertubation; sols++)
{
var timeleft = TotalTimelimit - (int)(DateTime.Now - starttime).TotalSeconds;
if (!model.Optimize(timeleft))
{
if (AbortSolverWhenNoImprovement)
{
break; //infeasibles
}
else
{
continue;
}
}
Console.WriteLine("new solution");
//constraint on objective function
// if (model.ObjSoftCons == prev) break;
currentObjective = model.ObjSoftCons;
model.SetQualConstraint(currentObjective);
var solututionAfter = new Solution(data, formulation);
solututionAfter.SetAssignments(model.GetAssignments());
Console.WriteLine($"Before: {before}\n" +
$"After: {model.ObjSoftCons}\n" +
$"Perbs: {pertubations}");
Console.WriteLine(solutionBefore.AnalyzeSolution());
Console.WriteLine(solututionAfter.AnalyzeSolution());
var solution = new Solution(data, formulation);
solution.SetAssignments(model.GetAssignments());
solution.AnalyzeSolution();
sol.Add(Tuple.Create(pertubations, solution, model.RunTime, (int)Math.Ceiling(model.ObjBound)));
DisplayPertubations(solutionBefore._assignments.ToList(), solututionAfter._assignments.ToList());
//to pertubate more.
model.RemoveCurrentSolution();
}
if (sols == 0)
{
break; //previous pertubations didnt find anything
}
if (AbortWhenPreviousSolutionValueIsObtained && currentObjective == (int)solutionBefore.Objective)
{
break;
}
if ((int)(DateTime.Now - starttime).TotalSeconds > TotalTimelimit)
{
Console.WriteLine("total timelimit reached");
break;
}
//overall timelimit?
model.SetQualConstraint(currentObjective - 1);
}
LastRuntimeSeconds = (int)(DateTime.Now - starttime).TotalSeconds;
return(sol);
}
// [TestCase(ITC_Comp05,1)]
/*
* [TestCase(ITC_Comp05, 2)]
* [TestCase(ITC_Comp05, 3)]
* [TestCase(ITC_Comp05, 4)]
* [TestCase(ITC_Comp05, 5)]
* [TestCase(ITC_Comp05, 6)]
* [TestCase(ITC_Comp05, 7)]
* [TestCase(ITC_Comp05, 8)]
* [TestCase(ITC_Comp05, 11)]
* [TestCase(ITC_Comp05, 12)]
* [TestCase(ITC_Comp05, 13)]
* [TestCase(ITC_Comp05, 14)]
* [TestCase(ITC_Comp05, 15)]
* [TestCase(ITC_Comp05, 16)]
* [TestCase(ITC_Comp05, 17)]
*/
public void CCfirstHeuristic30s(string filename, int seed) //
{
// var seed = 0;
for (int ia = 1; ia < 5; ia++)
{
seed = ia;
var data = Data.ReadXml(dataPath + filename, "120", "4");
// var newrooms = CreateRooms(data);
// data.SetRoomList(newrooms);
var problemFormulation = ProblemFormulation.UD2NoOverbook;
problemFormulation.MinimumWorkingDaysWeight = 0;
// problemFormulation.CurriculumCompactnessWeight = 0;
var par = new CCTModel.MIPModelParameters()
{
TuneGurobi = true,
UseStageIandII = false,
UseHallsConditions = true,
UseRoomHallConditions = true,
UseRoomsAsTypes = false,
Seed = seed,
};
var model = new CCTModel(data, problemFormulation, par);
model.SetObjective(1, 0, 0);
model.Optimize(TimelimitStageI);
model.DisplayObjectives();
var result = new Dictionary <double, double>();
//maybe find better proximity that reflects currciulumcompactness better.
//Check which curriculum that can be moved without introducting penalties.
model.SetProximityOrigin(true);
problemFormulation.MinimumWorkingDaysWeight = 5;
for (var i = 15; i <= 15; i++)
{
// model.SetObjective(0, 0, i);
model.SetProxConstraint(i);
model.AddMinimumWorkingDaysCost();
// model.FixCurricula(true);
model.Optimize();
model.DisplayObjectives();
model.SetObjective(1, 0, 0);
model.Fixsol(true);
model.Optimize();
model.Fixsol(false);
model.DisplayObjectives();
result[i] = model.Objective;
}
Console.WriteLine("Results:");
Console.WriteLine("i;obj");
foreach (var r in result)
{
Console.WriteLine($"{r.Key};{r.Value}");
}
var best = result.Min(r => r.Value);
File.AppendAllText(@"c:\temp\heuristic.txt", $"{nameof(CCfirstHeuristic30s)};{filename};{best}\n");
// model.FixCurricula(false);
// maybe also save multiple minimum curriculum solutions
}
}