static void Main(string[] args)
{
// read the arguments and stick them into the config manager
ConfigManager.Instance.ParseArguments(args);
// next need to get the file name and read the elements into dense training data
ITrainingData trainingData = new DenseTrainingData();
// initialize the dot product kernel, stick it in the SVM algorithm
// TODO: use the arguments to determine the kernel to use
// 1) dot product
// 2) radial basis function kernel
// 3) polynomial kernel (how to specify the degree of polynomial)
IKernel kernel = new DotProductKernel();
ISVMAlgorithm svm = new SMOSVMAlgorithm(trainingData, kernel);
// SVM algorithm run
CalculationStore data = svm.run();
// print the weight vectors and 'b' constants
data.Print();
// record the results of the SVM run
CalculationStore.RecordResults();
}
public void TrainingDataArgumentExceptionTest()
{
int lineIndex = 0;
IFileManager testFileManager = new SVM_SMO.Fakes.StubIFileManager()
{
ReadLine = () =>
{
if (lineIndex++ == 0) return "";
else return null;
},
ResetReader = () => { lineIndex = 0; }
};
ITrainingData trainingData = new DenseTrainingData(testFileManager);
}
public void TrainingDataTest()
{
int lineIndex = 0;
string[] lines = new [] { "0 1 2 1", "2 -4 1.5 -1", "" };
IFileManager testFileManager = new SVM_SMO.Fakes.StubIFileManager()
{
ReadLine = () => { return lines[lineIndex++]; },
ResetReader = () => { lineIndex = 0; }
};
ITrainingData trainingData = new DenseTrainingData(testFileManager);
Assert.AreEqual(0, trainingData.GetTrainingData(0, 0));
Assert.AreEqual(1, trainingData.GetTrainingData(0, 1));
Assert.AreEqual(2, trainingData.GetTrainingData(0, 2));
Assert.AreEqual(1, trainingData.GetTrainingResult(0));
Assert.AreEqual(2, trainingData.GetTrainingData(1, 0));
Assert.AreEqual(-4, trainingData.GetTrainingData(1, 1));
Assert.AreEqual(1.5, trainingData.GetTrainingData(1, 2));
Assert.AreEqual(-1, trainingData.GetTrainingResult(1));
}
