/// <summary>
/// Create a model from the training data, has to be called before predicting
/// </summary>
/// <param name="paramater">The SvmParameter is converted to OneClass type</param>
/// <returns>returns true if the model was created correctly</returns>
public bool CreateModel(SVMParameter parameter)
{
parameter.Type = SVMType.ONE_CLASS;
try
{
_model = _trainingData.Train(parameter);
return true;
}
catch
{
return false;
}
}
public static void Free(IntPtr ptr)
{
if (ptr == IntPtr.Zero)
{
return;
}
svm_model x = (svm_model)Marshal.PtrToStructure(ptr, typeof(svm_model));
SVMModel.Free(x);
Marshal.DestroyStructure(ptr, typeof(svm_model));
Marshal.FreeHGlobal(ptr);
ptr = IntPtr.Zero;
}
/// <summary>
///
/// </summary>
/// <param name="model"></param>
/// <param name="filename"></param>
/// <returns></returns>
public static bool SaveModel(SVMModel model, string filename)
{
if (String.IsNullOrWhiteSpace(filename) || model == null)
{
return(false);
}
IntPtr ptr_model = SVMModel.Allocate(model);
IntPtr ptr_filename = Marshal.StringToHGlobalAnsi(filename);
bool success = libsvm.svm_save_model(ptr_filename, ptr_model) == 0;
Marshal.FreeHGlobal(ptr_filename);
ptr_filename = IntPtr.Zero;
return(success);
}
/// <summary>
///
/// </summary>
/// <param name="model"></param>
/// <param name="x"></param>
/// <param name="estimations"></param>
/// <returns></returns>
public static double PredictProbability(SVMModel model, SVMNode[] x, out double[] estimations)
{
if (model == null)
{
throw new ArgumentNullException("model");
}
if (x == null)
{
throw new ArgumentNullException("x");
}
IntPtr ptr_model = SVMModel.Allocate(model);
double result = PredictProbability(ptr_model, x, out estimations);
SVMModel.Free(ptr_model);
return(result);
}
/// <summary>
/// This function does classification or regression on a test vector x given a model.
/// </summary>
/// <param name="model">SVM model.</param>
/// <param name="x">Test vector.</param>
/// <returns>For a classification model, the predicted class for x is returned.
/// For a regression model, the function value of x calculated using the model is returned.
/// For an one-class model, +1 or -1 is returned.</returns>
public static double Predict(SVMModel model, SVMNode[] x)
{
if (model == null)
{
throw new ArgumentNullException("model");
}
if (x == null)
{
throw new ArgumentNullException("x");
}
IntPtr ptr_model = SVMModel.Allocate(model);
double result = Predict(ptr_model, x);
SVMModel.Free(ptr_model);
return(result);
}
/// <summary>
///
/// </summary>
/// <param name="ptr_model"></param>
/// <param name="x"></param>
/// <param name="estimations"></param>
/// <returns></returns>
public static double PredictProbability(IntPtr ptr_model, SVMNode[] x, out double[] estimations)
{
if (ptr_model == IntPtr.Zero)
{
throw new ArgumentNullException("ptr_model");
}
if (x == null)
{
throw new ArgumentNullException("x");
}
bool isProbabilityModel = libsvm.svm_check_probability_model(ptr_model);
if (!isProbabilityModel)
{
SVMModel.Free(ptr_model);
estimations = null;
return(-1);
}
int classCount = libsvm.svm_get_nr_class(ptr_model);
IntPtr ptr_estimations = Marshal.AllocHGlobal(Marshal.SizeOf(typeof(double)) * classCount);
List <SVMNode> nodes = x.Select(a => a.Clone()).ToList();
nodes.Add(new SVMNode(-1, 0));
IntPtr ptr_nodes = SVMNode.Allocate(nodes.ToArray());
double result = libsvm.svm_predict_probability(ptr_model, ptr_nodes, ptr_estimations);
estimations = new double[classCount];
Marshal.Copy(ptr_estimations, estimations, 0, estimations.Length);
SVMNode.Free(ptr_nodes);
Marshal.FreeHGlobal(ptr_estimations);
ptr_estimations = IntPtr.Zero;
return(result);
}
/// <summary>
///
/// </summary>
/// <param name="model"></param>
/// <param name="x"></param>
/// <param name="values"></param>
/// <returns></returns>
public static double PredictValues(SVMModel model, SVMNode[] x, out double[] values)
{
int size = (int)(model.ClassCount * (model.ClassCount - 1) * 0.5);
IntPtr ptr_values = Marshal.AllocHGlobal(Marshal.SizeOf(typeof(double)) * size);
IntPtr ptr_model = SVMModel.Allocate(model);
List <SVMNode> nodes = x.Select(a => a.Clone()).ToList();
nodes.Add(new SVMNode(-1, 0));
IntPtr ptr_nodes = SVMNode.Allocate(nodes.ToArray());
double result = libsvm.svm_predict_values(ptr_model, ptr_nodes, ptr_values);
values = new double[size];
Marshal.Copy(ptr_values, values, 0, values.Length);
SVMModel.Free(ptr_model);
SVMNode.Free(ptr_nodes);
Marshal.FreeHGlobal(ptr_values);
ptr_values = IntPtr.Zero;
return(result);
}
public SVMModel Clone()
{
SVMModel y = new SVMModel();
if (Parameter != null)
y.Parameter = Parameter.Clone();
y.ClassCount = ClassCount;
y.TotalSVCount = TotalSVCount;
if (SV != null)
y.SV = SV.Select(a => a.Select(b => b.Clone()).ToArray()).ToList();
if (SVCoefs != null)
y.SVCoefs = SVCoefs.Select(a => a.Select(b => b).ToArray()).ToList();
if (Rho != null)
y.Rho = Rho.Select(a => a).ToArray();
if (ProbabilityA != null)
y.ProbabilityA = ProbabilityA.Select(a => a).ToArray();
if (ProbabilityB != null)
y.ProbabilityB = ProbabilityB.Select(a => a).ToArray();
if (SVIndices != null)
y.SVIndices = SVIndices.Select(a => a).ToArray();
if (Labels != null)
y.Labels = Labels.Select(a => a).ToArray();
if (SVCounts != null)
y.SVCounts = SVCounts.Select(a => a).ToArray();
y.Creation = Creation;
return y;
}
/// <summary>
/// This function constructs and returns an SVM model according to the given training data and parameters.
/// </summary>
/// <param name="problem">Training data.</param>
/// <param name="parameter">Parameter set.</param>
/// <returns>SVM model according to the given training data and parameters.</returns>
public static SVMModel Train(SVMProblem problem, SVMParameter parameter)
{
if (problem == null)
{
throw new ArgumentNullException("problem");
}
if (parameter == null)
{
throw new ArgumentNullException("parameter");
}
IntPtr ptr_problem = SVMProblem.Allocate(problem);
IntPtr ptr_parameter = SVMParameter.Allocate(parameter);
IntPtr ptr_model = libsvm.svm_train(ptr_problem, ptr_parameter);
SVMModel model = SVMModel.Convert(ptr_model);
SVMProblem.Free(ptr_problem);
SVMParameter.Free(ptr_parameter);
libsvm.svm_free_model_content(ptr_model);
return(model);
}
public static IntPtr Allocate(SVMModel x)
{
if (x == null || x.ClassCount < 1 || x.Parameter == null || x.Rho == null || x.Rho.Length < 1 ||
x.SVCoefs == null || x.SVCoefs.Count < 1 || x.TotalSVCount < 1)
{
return IntPtr.Zero;
}
if (x.Parameter.Type != SVMType.EPSILON_SVR && x.Parameter.Type != SVMType.NU_SVR &&
x.Parameter.Type != SVMType.ONE_CLASS &&
(x.Labels == null || x.Labels.Length < 1 || x.SVCounts == null || x.SVCounts.Length < 1))
{
return IntPtr.Zero;
}
svm_model y = new svm_model();
y.nr_class = x.ClassCount;
y.l = x.TotalSVCount;
y.free_sv = (int)x.Creation;
// Allocate model.parameter
IntPtr ptr_param = SVMParameter.Allocate(x.Parameter);
y.param = (svm_parameter)Marshal.PtrToStructure(ptr_param, typeof(svm_parameter));
SVMParameter.Free(ptr_param);
// Allocate model.rho
y.rho = Marshal.AllocHGlobal(Marshal.SizeOf(typeof(double)) * x.Rho.Length);
Marshal.Copy(x.Rho, 0, y.rho, x.Rho.Length);
// Allocate model.probA
y.probA = IntPtr.Zero;
if (x.ProbabilityA != null)
{
y.probA = Marshal.AllocHGlobal(Marshal.SizeOf(typeof(double)) * x.ProbabilityA.Length);
Marshal.Copy(x.ProbabilityA, 0, y.probA, x.ProbabilityA.Length);
}
// Allocate model.probB
y.probB = IntPtr.Zero;
if (x.ProbabilityB != null)
{
y.probB = Marshal.AllocHGlobal(Marshal.SizeOf(typeof(double)) * x.ProbabilityB.Length);
Marshal.Copy(x.ProbabilityB, 0, y.probB, x.ProbabilityB.Length);
}
if (x.Parameter.Type != SVMType.EPSILON_SVR && x.Parameter.Type != SVMType.NU_SVR &&
x.Parameter.Type != SVMType.ONE_CLASS)
{
// Allocate model.label
y.label = Marshal.AllocHGlobal(Marshal.SizeOf(typeof (int))*x.Labels.Length);
Marshal.Copy(x.Labels, 0, y.label, x.Labels.Length);
// Allocate model.nSV
y.nSV = Marshal.AllocHGlobal(Marshal.SizeOf(typeof (int))*x.SVCounts.Length);
Marshal.Copy(x.SVCounts, 0, y.nSV, x.SVCounts.Length);
}
else
{
y.label = IntPtr.Zero;
y.nSV = IntPtr.Zero;
}
// Allocate model.sv_coef
y.sv_coef = Marshal.AllocHGlobal(Marshal.SizeOf(typeof(IntPtr)) * x.SVCoefs.Count);
IntPtr i_ptr_svcoef = y.sv_coef;
for (int i = 0; i < x.SVCoefs.Count; i++)
{
IntPtr temp = Marshal.AllocHGlobal(Marshal.SizeOf(typeof(double)) * x.SVCoefs[i].Length);
Marshal.Copy(x.SVCoefs[i], 0, temp, x.SVCoefs[i].Length);
Marshal.StructureToPtr(temp, i_ptr_svcoef, true);
i_ptr_svcoef = IntPtr.Add(i_ptr_svcoef, Marshal.SizeOf(typeof(IntPtr)));
}
// Allocate model.sv_indices
y.sv_indices = IntPtr.Zero;
if (x.SVIndices != null)
{
y.sv_indices = Marshal.AllocHGlobal(Marshal.SizeOf(typeof(int)) * x.SVIndices.Length);
Marshal.Copy(x.SVIndices, 0, y.sv_indices, x.SVIndices.Length);
}
// Allocate model.SV
y.SV = IntPtr.Zero;
if (x.SV != null)
{
y.SV = Marshal.AllocHGlobal(Marshal.SizeOf(typeof(IntPtr)) * x.SV.Count);
IntPtr i_ptr_sv = y.SV;
for (int i = 0; i < x.SV.Count; i++)
{
// Prepare each node array
// 1) All nodes containing zero value is removed
// 2) A node which index is -1 is added to the end
List<SVMNode> temp = x.SV[i].Where(a => a.Value != 0).ToList();
temp.Add(new SVMNode(-1, 0));
SVMNode[] nodes = temp.ToArray();
// Allocate node array
IntPtr ptr_nodes = SVMNode.Allocate(nodes);
Marshal.StructureToPtr(ptr_nodes, i_ptr_sv, true);
i_ptr_sv = IntPtr.Add(i_ptr_sv, Marshal.SizeOf(typeof(IntPtr)));
}
}
// Allocate the model
int size = Marshal.SizeOf(y);
IntPtr ptr = Marshal.AllocHGlobal(size);
Marshal.StructureToPtr(y, ptr, true);
return ptr;
}
public static SVMModel Convert(svm_model x)
{
SVMModel y = new SVMModel();
y.Creation = (CreationType)x.free_sv;
y.ClassCount = x.nr_class;
y.TotalSVCount = x.l;
if (y.Creation == CreationType.LOAD_MODEL)
{
y.Parameter = new SVMParameter();
y.Parameter.Type = (SVMType)x.param.svm_type;
y.Parameter.Kernel = (SVMKernelType)x.param.kernel_type;
switch (y.Parameter.Kernel)
{
case SVMKernelType.LINEAR:
break;
case SVMKernelType.POLY:
y.Parameter.Gamma = x.param.gamma;
y.Parameter.Coef0 = x.param.coef0;
y.Parameter.Degree = x.param.degree;
break;
case SVMKernelType.RBF:
y.Parameter.Gamma = x.param.gamma;
break;
case SVMKernelType.SIGMOID:
y.Parameter.Gamma = x.param.gamma;
y.Parameter.Coef0 = x.param.coef0;
break;
}
}
else
{
y.Parameter = SVMParameter.Convert(x.param);
}
int problemCount = (int)(y.ClassCount * (y.ClassCount - 1) * 0.5);
y.Rho = new double[problemCount];
Marshal.Copy(x.rho, y.Rho, 0, y.Rho.Length);
y.ProbabilityA = null;
if (x.probA != IntPtr.Zero)
{
y.ProbabilityA = new double[problemCount];
Marshal.Copy(x.probA, y.ProbabilityA, 0, y.ProbabilityA.Length);
}
y.ProbabilityB = null;
if (x.probB != IntPtr.Zero)
{
y.ProbabilityB = new double[problemCount];
Marshal.Copy(x.probB, y.ProbabilityB, 0, y.ProbabilityB.Length);
}
if (x.nSV != IntPtr.Zero)
{
y.SVCounts = new int[y.ClassCount];
Marshal.Copy(x.nSV, y.SVCounts, 0, y.SVCounts.Length);
y.Labels = new int[y.ClassCount];
Marshal.Copy(x.label, y.Labels, 0, y.Labels.Length);
}
y.SVCoefs = new List<double[]>(y.ClassCount - 1);
IntPtr i_ptr_svcoef = x.sv_coef;
for (int i = 0; i < y.ClassCount - 1; i++)
{
y.SVCoefs.Add(new double[y.TotalSVCount]);
IntPtr coef_ptr = (IntPtr)Marshal.PtrToStructure(i_ptr_svcoef, typeof(IntPtr));
Marshal.Copy(coef_ptr, y.SVCoefs[i], 0, y.SVCoefs[i].Length);
i_ptr_svcoef = IntPtr.Add(i_ptr_svcoef, Marshal.SizeOf(typeof(IntPtr)));
}
y.SVIndices = null;
if (x.sv_indices != IntPtr.Zero)
{
y.SVIndices = new int[y.TotalSVCount];
Marshal.Copy(x.sv_indices, y.SVIndices, 0, y.SVIndices.Length);
}
y.SV = new List<SVMNode[]>();
IntPtr i_ptr_sv = x.SV;
for (int i = 0; i < x.l; i++)
{
IntPtr ptr_nodes = (IntPtr)Marshal.PtrToStructure(i_ptr_sv, typeof(IntPtr));
SVMNode[] nodes = SVMNode.Convert(ptr_nodes);
y.SV.Add(nodes);
i_ptr_sv = IntPtr.Add(i_ptr_sv, Marshal.SizeOf(typeof(IntPtr)));
}
return y;
}
/// <summary>
///
/// </summary>
/// <param name="model"></param>
/// <param name="filename"></param>
/// <returns></returns>
public static bool SaveModel(SVMModel model, string filename)
{
if (String.IsNullOrWhiteSpace(filename) || model == null)
{
return false;
}
IntPtr ptr_model = SVMModel.Allocate(model);
IntPtr ptr_filename = Marshal.StringToHGlobalAnsi(filename);
bool success = libsvm.svm_save_model(ptr_filename, ptr_model) == 0;
Marshal.FreeHGlobal(ptr_filename);
ptr_filename = IntPtr.Zero;
return success;
}
/// <summary>
///
/// </summary>
/// <param name="model"></param>
/// <param name="x"></param>
/// <param name="values"></param>
/// <returns></returns>
public static double PredictValues(SVMModel model, SVMNode[] x, out double[] values)
{
IntPtr ptr_model = SVMModel.Allocate(model);
double result = PredictValues(ptr_model, x, out values);
SVMModel.Free(ptr_model);
return result;
}
/// <summary>
///
/// </summary>
/// <param name="model"></param>
/// <param name="x"></param>
/// <param name="estimations"></param>
/// <returns></returns>
public static double PredictProbability(SVMModel model, SVMNode[] x, out double[] estimations)
{
IntPtr ptr_model = SVMModel.Allocate(model);
double result = PredictProbability(ptr_model, x, out estimations);
SVMModel.Free(ptr_model);
return result;
}
/// <summary>
/// This function does classification or regression on a test vector x given a model.
/// </summary>
/// <param name="model">SVM model.</param>
/// <param name="x">Test vector.</param>
/// <returns>For a classification model, the predicted class for x is returned.
/// For a regression model, the function value of x calculated using the model is returned.
/// For an one-class model, +1 or -1 is returned.</returns>
public static double Predict(SVMModel model, SVMNode[] x)
{
IntPtr ptr_model = SVMModel.Allocate(model);
double result = Predict(ptr_model, x);
SVMModel.Free(ptr_model);
return result;
}
/// <summary>
///
/// </summary>
/// <param name="model"></param>
/// <returns></returns>
public static bool CheckProbabilityModel(SVMModel model)
{
if (model == null)
{
throw new ArgumentNullException("model");
}
IntPtr ptr_model = SVMModel.Allocate(model);
bool success = libsvm.svm_check_probability_model(ptr_model);
SVMModel.Free(ptr_model);
return success;
}
public static SVMModel Convert(svm_model x)
{
SVMModel y = new SVMModel();
y.Creation = (CreationType)x.free_sv;
y.ClassCount = x.nr_class;
y.TotalSVCount = x.l;
if (y.Creation == CreationType.LOAD_MODEL)
{
y.Parameter = new SVMParameter();
y.Parameter.Type = (SVMType)x.param.svm_type;
y.Parameter.Kernel = (SVMKernelType)x.param.kernel_type;
switch (y.Parameter.Kernel)
{
case SVMKernelType.LINEAR:
break;
case SVMKernelType.POLY:
y.Parameter.Gamma = x.param.gamma;
y.Parameter.Coef0 = x.param.coef0;
y.Parameter.Degree = x.param.degree;
break;
case SVMKernelType.RBF:
y.Parameter.Gamma = x.param.gamma;
break;
case SVMKernelType.SIGMOID:
y.Parameter.Gamma = x.param.gamma;
y.Parameter.Coef0 = x.param.coef0;
break;
}
}
else
{
y.Parameter = SVMParameter.Convert(x.param);
}
int problemCount = (int)(y.ClassCount * (y.ClassCount - 1) * 0.5);
y.Rho = new double[problemCount];
Marshal.Copy(x.rho, y.Rho, 0, y.Rho.Length);
y.ProbabilityA = null;
if (x.probA != IntPtr.Zero)
{
y.ProbabilityA = new double[problemCount];
Marshal.Copy(x.probA, y.ProbabilityA, 0, y.ProbabilityA.Length);
}
y.ProbabilityB = null;
if (x.probB != IntPtr.Zero)
{
y.ProbabilityB = new double[problemCount];
Marshal.Copy(x.probB, y.ProbabilityB, 0, y.ProbabilityB.Length);
}
y.SVCounts = new int[y.ClassCount];
Marshal.Copy(x.nSV, y.SVCounts, 0, y.SVCounts.Length);
y.Labels = new int[y.ClassCount];
Marshal.Copy(x.label, y.Labels, 0, y.Labels.Length);
y.SVCoefs = new List <double[]>(y.ClassCount - 1);
IntPtr i_ptr_svcoef = x.sv_coef;
for (int i = 0; i < y.ClassCount - 1; i++)
{
y.SVCoefs.Add(new double[y.TotalSVCount]);
IntPtr coef_ptr = (IntPtr)Marshal.PtrToStructure(i_ptr_svcoef, typeof(IntPtr));
Marshal.Copy(coef_ptr, y.SVCoefs[i], 0, y.SVCoefs[i].Length);
i_ptr_svcoef = IntPtr.Add(i_ptr_svcoef, Marshal.SizeOf(typeof(IntPtr)));
}
y.SVIndices = null;
if (x.sv_indices != IntPtr.Zero)
{
y.SVIndices = new int[y.TotalSVCount];
Marshal.Copy(x.sv_indices, y.SVIndices, 0, y.SVIndices.Length);
}
y.SV = new List <SVMNode[]>();
IntPtr i_ptr_sv = x.SV;
for (int i = 0; i < x.l; i++)
{
IntPtr ptr_nodes = (IntPtr)Marshal.PtrToStructure(i_ptr_sv, typeof(IntPtr));
SVMNode[] nodes = SVMNode.Convert(ptr_nodes);
y.SV.Add(nodes);
i_ptr_sv = IntPtr.Add(i_ptr_sv, Marshal.SizeOf(typeof(IntPtr)));
}
return(y);
}
/// <summary>
/// This function does classification or regression on a test vector x given a model.
/// </summary>
/// <param name="model">SVM model.</param>
/// <param name="x">Test vector.</param>
/// <returns>For a classification model, the predicted class for x is returned.
/// For a regression model, the function value of x calculated using the model is returned.
/// For an one-class model, +1 or -1 is returned.</returns>
public static double Predict(SVMModel model, SVMNode[] x)
{
if (model == null)
{
throw new ArgumentNullException("model");
}
if (x == null)
{
throw new ArgumentNullException("x");
}
IntPtr ptr_model = SVMModel.Allocate(model);
double result = Predict(ptr_model, x);
SVMModel.Free(ptr_model);
return result;
}
/// <summary>
///
/// </summary>
/// <param name="model"></param>
/// <returns></returns>
public static bool CheckProbabilityModel(SVMModel model)
{
IntPtr ptr_model = SVMModel.Allocate(model);
bool success = libsvm.svm_check_probability_model(ptr_model);
SVMModel.Free(ptr_model);
return success;
}
/// <summary>
///
/// </summary>
/// <param name="model"></param>
/// <param name="x"></param>
/// <param name="estimations"></param>
/// <returns></returns>
public static double PredictProbability(SVMModel model, SVMNode[] x, out double[] estimations)
{
if (model == null)
{
throw new ArgumentNullException("model");
}
if (x == null)
{
throw new ArgumentNullException("x");
}
IntPtr ptr_model = SVMModel.Allocate(model);
double result = PredictProbability(ptr_model, x, out estimations);
SVMModel.Free(ptr_model);
return result;
}
public static IntPtr Allocate(SVMModel x)
{
if (x == null || x.ClassCount < 1 || x.Labels == null || x.Labels.Length < 1 || x.Parameter == null ||
x.Rho == null || x.Rho.Length < 1 || x.SVCoefs == null || x.SVCoefs.Count < 1 || x.TotalSVCount < 1 ||
x.SVCounts == null || x.SVCounts.Length < 1)
{
return(IntPtr.Zero);
}
svm_model y = new svm_model();
y.nr_class = x.ClassCount;
y.l = x.TotalSVCount;
y.free_sv = (int)x.Creation;
// Allocate model.parameter
IntPtr ptr_param = SVMParameter.Allocate(x.Parameter);
y.param = (svm_parameter)Marshal.PtrToStructure(ptr_param, typeof(svm_parameter));
// Allocate model.rho
y.rho = Marshal.AllocHGlobal(Marshal.SizeOf(typeof(double)) * x.Rho.Length);
Marshal.Copy(x.Rho, 0, y.rho, x.Rho.Length);
// Allocate model.probA
y.probA = IntPtr.Zero;
if (x.ProbabilityA != null)
{
y.probA = Marshal.AllocHGlobal(Marshal.SizeOf(typeof(double)) * x.ProbabilityA.Length);
Marshal.Copy(x.ProbabilityA, 0, y.probA, x.ProbabilityA.Length);
}
// Allocate model.probB
y.probB = IntPtr.Zero;
if (x.ProbabilityB != null)
{
y.probB = Marshal.AllocHGlobal(Marshal.SizeOf(typeof(double)) * x.ProbabilityB.Length);
Marshal.Copy(x.ProbabilityB, 0, y.probB, x.ProbabilityB.Length);
}
// Allocate model.label
y.label = Marshal.AllocHGlobal(Marshal.SizeOf(typeof(int)) * x.Labels.Length);
Marshal.Copy(x.Labels, 0, y.label, x.Labels.Length);
// Allocate model.nSV
y.nSV = Marshal.AllocHGlobal(Marshal.SizeOf(typeof(int)) * x.SVCounts.Length);
Marshal.Copy(x.SVCounts, 0, y.nSV, x.SVCounts.Length);
// Allocate model.sv_coef
y.sv_coef = Marshal.AllocHGlobal(Marshal.SizeOf(typeof(IntPtr)) * x.SVCoefs.Count);
IntPtr i_ptr_svcoef = y.sv_coef;
for (int i = 0; i < x.SVCoefs.Count; i++)
{
IntPtr temp = Marshal.AllocHGlobal(Marshal.SizeOf(typeof(double)) * x.SVCoefs[i].Length);
Marshal.Copy(x.SVCoefs[i], 0, temp, x.SVCoefs[i].Length);
Marshal.StructureToPtr(temp, i_ptr_svcoef, true);
i_ptr_svcoef = IntPtr.Add(i_ptr_svcoef, Marshal.SizeOf(typeof(IntPtr)));
}
// Allocate model.sv_indices
y.sv_indices = IntPtr.Zero;
if (x.SVIndices != null)
{
y.sv_indices = Marshal.AllocHGlobal(Marshal.SizeOf(typeof(int)) * x.SVIndices.Length);
Marshal.Copy(x.SVIndices, 0, y.sv_indices, x.SVIndices.Length);
}
// Allocate model.SV
y.SV = IntPtr.Zero;
if (x.SV != null)
{
y.SV = Marshal.AllocHGlobal(Marshal.SizeOf(typeof(IntPtr)) * x.SV.Count);
IntPtr i_ptr_sv = y.SV;
for (int i = 0; i < x.SV.Count; i++)
{
// Prepare each node array
// 1) All nodes containing zero value is removed
// 2) A node which index is -1 is added to the end
List <SVMNode> temp = x.SV[i].Where(a => a.Value != 0).ToList();
temp.Add(new SVMNode(-1, 0));
SVMNode[] nodes = temp.ToArray();
// Allocate node array
IntPtr ptr_nodes = SVMNode.Allocate(nodes);
Marshal.StructureToPtr(ptr_nodes, i_ptr_sv, true);
i_ptr_sv = IntPtr.Add(i_ptr_sv, Marshal.SizeOf(typeof(IntPtr)));
}
}
// Allocate the model
int size = Marshal.SizeOf(y);
IntPtr ptr = Marshal.AllocHGlobal(size);
Marshal.StructureToPtr(y, ptr, true);
return(ptr);
}