public void Train()
{
int datasetSize = _dataMgr.count;
int dimension = _dataMgr.inputNum;
Node[][] inputs = new Node[datasetSize][];
double[] labels = new double[datasetSize];
int i = 0;
int j = 0;
for (i = 0; i < datasetSize; i++)
{
int index = i; // _rnd.Next(datasetSize - 1);
inputs[i] = new Node[dimension];
double[] input = _dataMgr.GetInputData(index);
for (j = 0; j < dimension; j++)
{
inputs[i][j] = new Node();
inputs[i][j].Index = j;
inputs[i][j].Value = input[j];
}
if (_dataMgr.GetLabelData(index) < 0.5)
{
labels[i] = -1;
}
else
{
labels[i] = 1;
}
}
Problem problem = new Problem(datasetSize, labels, inputs, dimension);
//RangeTransform range = RangeTransform.Compute(problem);
//problem = range.Scale(problem);
Parameter param = new Parameter();
param.C = 2;
param.Gamma = .5;
//param.KernelType = KernelType.RBF;
//param.SvmType = SvmType.C_SVC;
_model = new Model();
_model.Parameter = param;
bool[] nonzero = new bool[datasetSize];
double[] alpha = new double[problem.Count];
int l = problem.Count;
double[] Minus_ones = new double[l];
sbyte[] y = new sbyte[l];
for (i = 0; i < l; i++)
{
alpha[i] = 0;
Minus_ones[i] = -1;
if (problem.Y[i] > 0)
{
y[i] = +1;
}
else
{
y[i] = -1;
}
}
Solver s = new Solver();
Solver.SolutionInfo si = new Solver.SolutionInfo();
s.Solve(l, new SVC_Q(problem, param, y), Minus_ones, y,
alpha, param.C, param.C, param.EPS, si, param.Shrinking);
double sum_alpha = 0;
for (i = 0; i < l; i++)
{
sum_alpha += alpha[i];
}
for (i = 0; i < l; i++)
{
alpha[i] *= y[i];
}
Procedures.decision_function f = new Procedures.decision_function();
f.alpha = alpha;
f.rho = si.rho;
// build output
_model.NumberOfClasses = 2;
_model.ClassLabels = new int[2];
_model.ClassLabels[0] = 1;
_model.ClassLabels[1] = -1;
_model.Rho = new double[1];
_model.Rho[0] = f.rho;
_model.NumberOfSVPerClass = new int[2];
int posCnt = 0, negCnt = 0;
for (i = 0; i < datasetSize; i++)
{
if (labels[i] > 0)
{
posCnt++;
}
else
{
negCnt++;
}
}
_model.NumberOfSVPerClass[0] = posCnt;
_model.NumberOfSVPerClass[1] = negCnt;
_model.SupportVectorCount = datasetSize;
_model.SupportVectors = new Node[datasetSize][];
_model.SupportVectorCoefficients = new double[1][];
_model.SupportVectorCoefficients[0] = new double[datasetSize];
int p = 0;
for (i = 0; i < datasetSize; i++)
{
_model.SupportVectors[p++] = inputs[i];
_model.SupportVectorCoefficients[0][i] = f.alpha[i];
}
//_model = Training.Train(problem, param);
}
