protected double Run(StreamWriter log, string inputFile, int timeLimit)
{
int startTime, elapsedTime;
string outputFile = Path.GetTempFileName();
int maxTime = timeLimit + (int)Math.Round((TimeTolerance / 100.0) * timeLimit);
startTime = Environment.TickCount;
Metaheuristic.Start(inputFile, outputFile, timeLimit);
elapsedTime = Environment.TickCount - startTime;
log.WriteLine("Elapsed time for " + Path.GetFileName(inputFile) + ": " + elapsedTime + " of " + maxTime);
if (elapsedTime > maxTime)
{
return(double.PositiveInfinity);
}
else
{
using (StreamReader reader = File.OpenText(outputFile)) {
string line = reader.ReadLine();
while (line.Trim() == "")
{
line = reader.ReadLine();
}
return(double.Parse(line));
}
}
}
public void Start(string logFile)
{
StreamWriter log = File.CreateText(logFile);
double[] bestParameters = null;
double bestCost = double.PositiveInfinity;
double currentCost;
double currentFitness;
List <double> meanFitness = new List <double>();
List <double> runFitness = new List <double>();
log.AutoFlush = true;
log.WriteLine("Parameters tuning for: " + Metaheuristic.Name);
log.WriteLine("Estimated run time: " + ApproxRunningTime());
foreach (double[] currentParameters in EnumerateParameters())
{
log.WriteLine();
log.WriteLine("Current parameters: " + ParametersToString(currentParameters));
Metaheuristic.UpdateParameters(currentParameters);
meanFitness.Clear();
foreach (string inputFile in InputFiles)
{
runFitness.Clear();
foreach (int timeLimit in TimeLimits)
{
for (int run = 0; run < RunsPerProblem; run++)
{
currentFitness = Run(log, inputFile, timeLimit);
if (double.IsPositiveInfinity(currentFitness))
{
runFitness.Clear();
break;
}
runFitness.Add(currentFitness);
}
if (runFitness.Count == 0)
{
meanFitness.Clear();
break;
}
}
if (runFitness.Count == 0)
{
meanFitness.Clear();
break;
}
meanFitness.Add(Statistics.Mean(runFitness));
log.WriteLine("Mean fitness for " + Path.GetFileName(inputFile) + ": " + meanFitness[meanFitness.Count - 1]);
}
if (meanFitness.Count == 0)
{
log.WriteLine("Skiping this parameters.");
continue;
}
currentCost = meanFitness.Sum();
log.WriteLine("Current cost: " + currentCost);
if (currentCost < bestCost)
{
bestCost = currentCost;
bestParameters = currentParameters;
log.WriteLine("This is the best cost until now.");
}
}
log.WriteLine();
if (bestParameters != null)
{
log.WriteLine("Final best parameters: " + ParametersToString(bestParameters));
}
else
{
log.WriteLine("Not feasible parameters found.");
}
log.Close();
}
