public ProblemInstance Produce(TrademarkRepository rep)
{
var random = new Random();
var res = new ProblemInstance {
TrademarkRepository = rep
};
int n = random.Next(2, rep.AllTrademarks.Count);
var trademarkIndices = Enumerable.Range(0, rep.AllTrademarks.Count).ToArray();
Shuffle(trademarkIndices);
for (int i = 0; i < n; i++)
{
if (res.TrainingSetRatio <= trainingSetRatio)
{
res.TrainingSet.Add(rep.AllTrademarks[trademarkIndices[i]]);
}
else
{
res.ControlSet.Add(rep.AllTrademarks[trademarkIndices[i]]);
}
}
return(res);
}
public ProblemInstance Produce(TrademarkRepository rep)
{
var res = new ProblemInstance {
TrademarkRepository = rep
};
foreach (var trademark in rep.AllTrademarks)
{
if (Path.GetFileNameWithoutExtension(trademark.FilePath).Contains("T"))
{
res.TrainingSet.Add(trademark);
}
else
{
res.ControlSet.Add(trademark);
}
}
return(res);
}
static void Main()
{
var preprocessor = new AForgeImagePreprocessor();
var rep = new TrademarkRepository(InputDir, preprocessor);
var classifier = new GansClassifier(rep.AllTrademarkClasses.Count);
//var classifier = new ColorMomentANNClassifier(rep.AllTrademarkClasses.Count);
//var classifier = new ColorMomentMetricClassifier();
var estimator = new CrossValidationEstimator(40, TrainingSetRatio, rep);
//var estimator = new NaiveEstimator(problem);
Console.WriteLine("Classes: {0} ({1:P02} each); Training set: {2:F0}; Control set: {3:F0}", rep.AllTrademarkClasses.Count, 1.0 / rep.AllTrademarkClasses.Count, rep.AllTrademarks.Count * TrainingSetRatio, rep.AllTrademarks.Count * (1 - TrainingSetRatio));
var estimate = estimator.Estimate(classifier);
Console.WriteLine(@"{0}: {1:P02}", classifier.GetType().Name, estimate);
}
public CrossValidationEstimator(int sampleCount, double trainingSetRatio, TrademarkRepository trademarkRepository)
{
this.sampleCount = sampleCount;
this.trademarkRepository = trademarkRepository;
problemSource = new RandomProblemSource(trainingSetRatio);
}
