/// <summary>
/// Performs cross validation.
/// </summary>
/// <param name="problem">The training data</param>
/// <param name="parameters">The parameters to test</param>
/// <param name="nrfold">The number of cross validations to use</param>
/// <returns>The cross validation score</returns>
public static double PerformCrossValidation(Problem problem, Parameter parameters, int nrfold)
{
string error = Procedures.svm_check_parameter(problem, parameters);
if (error == null)
return doCrossValidation(problem, parameters, nrfold);
else throw new Exception(error);
}
public static void Main(string[] args)
{
Problem train = Problem.Read("a1a.train.txt");
Problem test = Problem.Read("a1a.test.txt");
//For this example (and indeed, many scenarios), the default
//parameters will suffice.
Parameter parameters = new Parameter();
double C;
double Gamma;
//This will do a grid optimization to find the best parameters
//and store them in C and Gamma, outputting the entire
//search to params.txt.
ParameterSelection.Grid(train, parameters, "params.txt", out C, out Gamma);
parameters.C = C;
parameters.Gamma = Gamma;
//Train the model using the optimal parameters.
Model model = Training.Train(train, parameters);
//Perform classification on the test data, putting the
//results in results.txt.
Prediction.Predict(test, "results.txt", model, false);
}
/// <summary>
/// Constructor.
/// </summary>
/// <param name="rows">Nodes to use as the rows of the matrix</param>
/// <param name="columns">Nodes to use as the columns of the matrix</param>
/// <param name="param">Parameters to use when compute similarities</param>
public PrecomputedKernel(List<Node[]> rows, List<Node[]> columns, Parameter param)
{
_rows = rows.Count;
_columns = columns.Count;
_similarities = new float[_rows, _columns];
for (int r = 0; r < _rows; r++)
for (int c = 0; c < _columns; c++)
_similarities[r, c] = (float)Kernel.KernelFunction(rows[r], columns[c], param);
}
/// <summary>
/// Constructor.
/// </summary>
/// <param name="nodes">Nodes for self-similarity analysis</param>
/// <param name="param">Parameters to use when computing similarities</param>
public PrecomputedKernel(List<Node[]> nodes, Parameter param)
{
_rows = nodes.Count;
_columns = _rows;
_similarities = new float[_rows, _columns];
for (int r = 0; r < _rows; r++)
{
for (int c = 0; c < r; c++)
_similarities[r, c] = _similarities[c, r];
_similarities[r, r] = 1;
for (int c = r + 1; c < _columns; c++)
_similarities[r, c] = (float)Kernel.KernelFunction(nodes[r], nodes[c], param);
}
}
public Kernel(int l, Node[][] x_, Parameter param)
{
_kernelType = param.KernelType;
_degree = param.Degree;
_gamma = param.Gamma;
_coef0 = param.Coefficient0;
_x = (Node[][])x_.Clone();
if (_kernelType == KernelType.RBF) {
_xSquare = new double[l];
for (int i = 0; i < l; i++)
_xSquare[i] = dot(_x[i], _x[i]);
} else _xSquare = null;
}
public Model train(Problem issue)
{
var span = Overseer.observe("Training.Parameter-Choosing");
Parameter parameters = new Parameter();
parameters.KernelType = KernelType.RBF;
double C;
double Gamma;
ParameterSelection.Grid(issue, parameters, null, out C, out Gamma);
parameters.C = C;
parameters.Gamma = Gamma;
span.die();
span = Overseer.observe("Training.Training");
var result = Training.Train(issue, parameters);
span.die();
return result;
}
public static double KernelFunction(Node[] x, Node[] y, Parameter param)
{
switch (param.KernelType) {
case KernelType.LINEAR:
return dot(x, y);
case KernelType.POLY:
return powi(param.Degree * dot(x, y) + param.Coefficient0, param.Degree);
case KernelType.RBF: {
double sum = computeSquaredDistance(x, y);
return Math.Exp(-param.Gamma * sum);
}
case KernelType.SIGMOID:
return Math.Tanh(param.Gamma * dot(x, y) + param.Coefficient0);
case KernelType.PRECOMPUTED:
return x[(int)(y[0].Value)].Value;
default:
return 0;
}
}
///
public override void LearnAttributeToFactorMapping()
{
var svm_features = new List<Node[]>();
var relevant_items = new List<int>();
for (int i = 0; i < MaxItemID + 1; i++)
{
// ignore items w/o collaborative data
if (Feedback.ItemMatrix[i].Count == 0)
continue;
// ignore items w/o attribute data
if (item_attributes[i].Count == 0)
continue;
svm_features.Add( CreateNodes(i) );
relevant_items.Add(i);
}
// TODO proper random seed initialization
Node[][] svm_features_array = svm_features.ToArray();
var svm_parameters = new Parameter();
svm_parameters.SvmType = SvmType.EPSILON_SVR;
//svm_parameters.SvmType = SvmType.NU_SVR;
svm_parameters.C = this.c;
svm_parameters.Gamma = this.gamma;
models = new Model[num_factors];
for (int f = 0; f < num_factors; f++)
{
double[] targets = new double[svm_features.Count];
for (int i = 0; i < svm_features.Count; i++)
{
int item_id = relevant_items[i];
targets[i] = item_factors[item_id, f];
}
Problem svm_problem = new Problem(svm_features.Count, targets, svm_features_array, NumItemAttributes - 1);
models[f] = SVM.Training.Train(svm_problem, svm_parameters);
}
_MapToLatentFactorSpace = Utils.Memoize<int, double[]>(__MapToLatentFactorSpace);
}
private double TrainAndTest(string trainSet,string testSet, string resultFile)
{
Problem train = Problem.Read(trainSet);
Problem test = Problem.Read(testSet);
Parameter parameters = new Parameter();
if (chClassification.Checked)
{
parameters.SvmType = SvmType.C_SVC;
parameters.C = 0.03;
parameters.Gamma = 0.008;
}
else
{
parameters.SvmType = SvmType.EPSILON_SVR;
parameters.C = 8;
parameters.Gamma = 0.063;
parameters.P = 0.5;
}
Model model = Training.Train(train, parameters);
return Prediction.Predict(test, resultFile, model, true);
}
public ONE_CLASS_Q(Problem prob, Parameter param)
: base(prob.Count, prob.X, param)
{
cache = new Cache(prob.Count, (long)(param.CacheSize * (1 << 20)));
QD = new float[prob.Count];
for (int i = 0; i < prob.Count; i++)
QD[i] = (float)KernelFunction(i, i);
}
// Stratified cross validation
public static void svm_cross_validation(Problem prob, Parameter param, int nr_fold, double[] target)
{
Random rand = new Random();
int i;
int[] fold_start = new int[nr_fold + 1];
int l = prob.Count;
int[] perm = new int[l];
// stratified cv may not give leave-one-out rate
// Each class to l folds -> some folds may have zero elements
if ((param.SvmType == SvmType.C_SVC ||
param.SvmType == SvmType.NU_SVC) && nr_fold < l)
{
int[] tmp_nr_class = new int[1];
int[][] tmp_label = new int[1][];
int[][] tmp_start = new int[1][];
int[][] tmp_count = new int[1][];
svm_group_classes(prob, tmp_nr_class, tmp_label, tmp_start, tmp_count, perm);
int nr_class = tmp_nr_class[0];
//int[] label = tmp_label[0];
int[] start = tmp_start[0];
int[] count = tmp_count[0];
// random shuffle and then data grouped by fold using the array perm
int[] fold_count = new int[nr_fold];
int c;
int[] index = new int[l];
for (i = 0; i < l; i++)
index[i] = perm[i];
for (c = 0; c < nr_class; c++)
for (i = 0; i < count[c]; i++)
{
int j = i + (int)(rand.NextDouble() * (count[c] - i));
do { int _ = index[start[c] + j]; index[start[c] + j] = index[start[c] + i]; index[start[c] + i] = _; } while (false);
}
for (i = 0; i < nr_fold; i++)
{
fold_count[i] = 0;
for (c = 0; c < nr_class; c++)
fold_count[i] += (i + 1) * count[c] / nr_fold - i * count[c] / nr_fold;
}
fold_start[0] = 0;
for (i = 1; i <= nr_fold; i++)
fold_start[i] = fold_start[i - 1] + fold_count[i - 1];
for (c = 0; c < nr_class; c++)
for (i = 0; i < nr_fold; i++)
{
int begin = start[c] + i * count[c] / nr_fold;
int end = start[c] + (i + 1) * count[c] / nr_fold;
for (int j = begin; j < end; j++)
{
perm[fold_start[i]] = index[j];
fold_start[i]++;
}
}
fold_start[0] = 0;
for (i = 1; i <= nr_fold; i++)
fold_start[i] = fold_start[i - 1] + fold_count[i - 1];
}
else
{
for (i = 0; i < l; i++) perm[i] = i;
for (i = 0; i < l; i++)
{
int j = i + (int)(rand.NextDouble() * (l - i));
do { int _ = perm[i]; perm[i] = perm[j]; perm[j] = _; } while (false);
}
for (i = 0; i <= nr_fold; i++)
fold_start[i] = i * l / nr_fold;
}
for (i = 0; i < nr_fold; i++)
{
int begin = fold_start[i];
int end = fold_start[i + 1];
int j, k;
Problem subprob = new Problem();
subprob.Count = l - (end - begin);
subprob.X = new Node[subprob.Count][];
subprob.Y = new double[subprob.Count];
k = 0;
for (j = 0; j < begin; j++)
{
subprob.X[k] = prob.X[perm[j]];
subprob.Y[k] = prob.Y[perm[j]];
++k;
}
for (j = end; j < l; j++)
{
subprob.X[k] = prob.X[perm[j]];
subprob.Y[k] = prob.Y[perm[j]];
++k;
}
Model submodel = svm_train(subprob, param);
if (param.Probability &&
(param.SvmType == SvmType.C_SVC ||
param.SvmType == SvmType.NU_SVC))
{
double[] prob_estimates = new double[svm_get_nr_class(submodel)];
for (j = begin; j < end; j++)
target[perm[j]] = svm_predict_probability(submodel, prob.X[perm[j]], prob_estimates);
}
else
for (j = begin; j < end; j++)
target[perm[j]] = svm_predict(submodel, prob.X[perm[j]]);
}
}
static decision_function svm_train_one(Problem prob, Parameter param, double Cp, double Cn)
{
double[] alpha = new double[prob.Count];
Solver.SolutionInfo si = new Solver.SolutionInfo();
switch (param.SvmType)
{
case SvmType.C_SVC:
solve_c_svc(prob, param, alpha, si, Cp, Cn);
break;
case SvmType.NU_SVC:
solve_nu_svc(prob, param, alpha, si);
break;
case SvmType.ONE_CLASS:
solve_one_class(prob, param, alpha, si);
break;
case SvmType.EPSILON_SVR:
solve_epsilon_svr(prob, param, alpha, si);
break;
case SvmType.NU_SVR:
solve_nu_svr(prob, param, alpha, si);
break;
}
Procedures.info("obj = " + si.obj + ", rho = " + si.rho + "\n");
// output SVs
int nSV = 0;
int nBSV = 0;
for (int i = 0; i < prob.Count; i++)
{
if (Math.Abs(alpha[i]) > 0)
{
++nSV;
if (prob.Y[i] > 0)
{
if (Math.Abs(alpha[i]) >= si.upper_bound_p)
++nBSV;
}
else
{
if (Math.Abs(alpha[i]) >= si.upper_bound_n)
++nBSV;
}
}
}
Procedures.info("nSV = " + nSV + ", nBSV = " + nBSV + "\n");
decision_function f = new decision_function();
f.alpha = alpha;
f.rho = si.rho;
return f;
}
/// <summary>
/// Performs a Grid parameter selection, trying all possible combinations of the two lists and returning the
/// combination which performed best. Uses the default values of C and Gamma.
/// </summary>
/// <param name="problem">The training data</param>
/// <param name="validation">The validation data</param>
/// <param name="parameters">The parameters to use when optimizing</param>
/// <param name="outputFile">The output file for the parameter results</param>
/// <param name="C">The optimal C value will be placed in this variable</param>
/// <param name="Gamma">The optimal Gamma value will be placed in this variable</param>
public static void Grid(
Problem problem,
Problem validation,
Parameter parameters,
string outputFile,
out double C,
out double Gamma)
{
Grid(problem, validation, parameters, GetList(MIN_C, MAX_C, C_STEP), GetList(MIN_G, MAX_G, G_STEP), outputFile, out C, out Gamma);
}
private static void solve_nu_svc(Problem prob, Parameter param,
double[] alpha, Solver.SolutionInfo si)
{
int i;
int l = prob.Count;
double nu = param.Nu;
sbyte[] y = new sbyte[l];
for (i = 0; i < l; i++)
if (prob.Y[i] > 0)
y[i] = +1;
else
y[i] = -1;
double sum_pos = nu * l / 2;
double sum_neg = nu * l / 2;
for (i = 0; i < l; i++)
if (y[i] == +1)
{
alpha[i] = Math.Min(1.0, sum_pos);
sum_pos -= alpha[i];
}
else
{
alpha[i] = Math.Min(1.0, sum_neg);
sum_neg -= alpha[i];
}
double[] zeros = new double[l];
for (i = 0; i < l; i++)
zeros[i] = 0;
Solver_NU s = new Solver_NU();
s.Solve(l, new SVC_Q(prob, param, y), zeros, y, alpha, 1.0, 1.0, param.EPS, si, param.Shrinking);
double r = si.r;
Procedures.info("C = " + 1 / r + "\n");
for (i = 0; i < l; i++)
alpha[i] *= y[i] / r;
si.rho /= r;
si.obj /= (r * r);
si.upper_bound_p = 1 / r;
si.upper_bound_n = 1 / r;
}
private static void solve_c_svc(Problem prob, Parameter param,
double[] alpha, Solver.SolutionInfo si,
double Cp, double Cn)
{
int l = prob.Count;
double[] Minus_ones = new double[l];
sbyte[] y = new sbyte[l];
int i;
for (i = 0; i < l; i++)
{
alpha[i] = 0;
Minus_ones[i] = -1;
if (prob.Y[i] > 0) y[i] = +1; else y[i] = -1;
}
Solver s = new Solver();
s.Solve(l, new SVC_Q(prob, param, y), Minus_ones, y,
alpha, Cp, Cn, param.EPS, si, param.Shrinking);
double sum_alpha = 0;
for (i = 0; i < l; i++)
sum_alpha += alpha[i];
if (Cp == Cn)
Procedures.info("nu = " + sum_alpha / (Cp * prob.Count) + "\n");
for (i = 0; i < l; i++)
alpha[i] *= y[i];
}
//
// Interface functions
//
public static Model svm_train(Problem prob, Parameter param)
{
Model model = new Model();
model.Parameter = param;
if (param.SvmType == SvmType.ONE_CLASS ||
param.SvmType == SvmType.EPSILON_SVR ||
param.SvmType == SvmType.NU_SVR)
{
// regression or one-class-svm
model.NumberOfClasses = 2;
model.ClassLabels = null;
model.NumberOfSVPerClass = null;
model.PairwiseProbabilityA = null; model.PairwiseProbabilityB = null;
model.SupportVectorCoefficients = new double[1][];
if (param.Probability &&
(param.SvmType == SvmType.EPSILON_SVR ||
param.SvmType == SvmType.NU_SVR))
{
model.PairwiseProbabilityA = new double[1];
model.PairwiseProbabilityA[0] = svm_svr_probability(prob, param);
}
decision_function f = svm_train_one(prob, param, 0, 0);
model.Rho = new double[1];
model.Rho[0] = f.rho;
int nSV = 0;
int i;
for (i = 0; i < prob.Count; i++)
if (Math.Abs(f.alpha[i]) > 0) ++nSV;
model.SupportVectorCount = nSV;
model.SupportVectors = new Node[nSV][];
model.SupportVectorCoefficients[0] = new double[nSV];
int j = 0;
for (i = 0; i < prob.Count; i++)
if (Math.Abs(f.alpha[i]) > 0)
{
model.SupportVectors[j] = prob.X[i];
model.SupportVectorCoefficients[0][j] = f.alpha[i];
++j;
}
}
else
{
// classification
int l = prob.Count;
int[] tmp_nr_class = new int[1];
int[][] tmp_label = new int[1][];
int[][] tmp_start = new int[1][];
int[][] tmp_count = new int[1][];
int[] perm = new int[l];
// group training data of the same class
svm_group_classes(prob, tmp_nr_class, tmp_label, tmp_start, tmp_count, perm);
int nr_class = tmp_nr_class[0];
int[] label = tmp_label[0];
int[] start = tmp_start[0];
int[] count = tmp_count[0];
Node[][] x = new Node[l][];
int i;
for (i = 0; i < l; i++)
x[i] = prob.X[perm[i]];
// calculate weighted C
double[] weighted_C = new double[nr_class];
for (i = 0; i < nr_class; i++)
weighted_C[i] = param.C;
foreach (int weightedLabel in param.Weights.Keys)
{
int index = Array.IndexOf<int>(label, weightedLabel);
if (index < 0)
Console.Error.WriteLine("warning: class label " + weightedLabel + " specified in weight is not found");
else weighted_C[index] *= param.Weights[weightedLabel];
}
// train k*(k-1)/2 models
bool[] nonzero = new bool[l];
for (i = 0; i < l; i++)
nonzero[i] = false;
decision_function[] f = new decision_function[nr_class * (nr_class - 1) / 2];
double[] probA = null, probB = null;
if (param.Probability)
{
probA = new double[nr_class * (nr_class - 1) / 2];
probB = new double[nr_class * (nr_class - 1) / 2];
}
int p = 0;
for (i = 0; i < nr_class; i++)
for (int j = i + 1; j < nr_class; j++)
{
Problem sub_prob = new Problem();
int si = start[i], sj = start[j];
int ci = count[i], cj = count[j];
sub_prob.Count = ci + cj;
sub_prob.X = new Node[sub_prob.Count][];
sub_prob.Y = new double[sub_prob.Count];
int k;
for (k = 0; k < ci; k++)
{
sub_prob.X[k] = x[si + k];
sub_prob.Y[k] = +1;
}
for (k = 0; k < cj; k++)
{
sub_prob.X[ci + k] = x[sj + k];
sub_prob.Y[ci + k] = -1;
}
if (param.Probability)
{
double[] probAB = new double[2];
svm_binary_svc_probability(sub_prob, param, weighted_C[i], weighted_C[j], probAB);
probA[p] = probAB[0];
probB[p] = probAB[1];
}
f[p] = svm_train_one(sub_prob, param, weighted_C[i], weighted_C[j]);
for (k = 0; k < ci; k++)
if (!nonzero[si + k] && Math.Abs(f[p].alpha[k]) > 0)
nonzero[si + k] = true;
for (k = 0; k < cj; k++)
if (!nonzero[sj + k] && Math.Abs(f[p].alpha[ci + k]) > 0)
nonzero[sj + k] = true;
++p;
}
// build output
model.NumberOfClasses = nr_class;
model.ClassLabels = new int[nr_class];
for (i = 0; i < nr_class; i++)
model.ClassLabels[i] = label[i];
model.Rho = new double[nr_class * (nr_class - 1) / 2];
for (i = 0; i < nr_class * (nr_class - 1) / 2; i++)
model.Rho[i] = f[i].rho;
if (param.Probability)
{
model.PairwiseProbabilityA = new double[nr_class * (nr_class - 1) / 2];
model.PairwiseProbabilityB = new double[nr_class * (nr_class - 1) / 2];
for (i = 0; i < nr_class * (nr_class - 1) / 2; i++)
{
model.PairwiseProbabilityA[i] = probA[i];
model.PairwiseProbabilityB[i] = probB[i];
}
}
else
{
model.PairwiseProbabilityA = null;
model.PairwiseProbabilityB = null;
}
int nnz = 0;
int[] nz_count = new int[nr_class];
model.NumberOfSVPerClass = new int[nr_class];
for (i = 0; i < nr_class; i++)
{
int nSV = 0;
for (int j = 0; j < count[i]; j++)
if (nonzero[start[i] + j])
{
++nSV;
++nnz;
}
model.NumberOfSVPerClass[i] = nSV;
nz_count[i] = nSV;
}
Procedures.info("Total nSV = " + nnz + "\n");
model.SupportVectorCount = nnz;
model.SupportVectors = new Node[nnz][];
p = 0;
for (i = 0; i < l; i++)
if (nonzero[i]) model.SupportVectors[p++] = x[i];
int[] nz_start = new int[nr_class];
nz_start[0] = 0;
for (i = 1; i < nr_class; i++)
nz_start[i] = nz_start[i - 1] + nz_count[i - 1];
model.SupportVectorCoefficients = new double[nr_class - 1][];
for (i = 0; i < nr_class - 1; i++)
model.SupportVectorCoefficients[i] = new double[nnz];
p = 0;
for (i = 0; i < nr_class; i++)
for (int j = i + 1; j < nr_class; j++)
{
// classifier (i,j): coefficients with
// i are in sv_coef[j-1][nz_start[i]...],
// j are in sv_coef[i][nz_start[j]...]
int si = start[i];
int sj = start[j];
int ci = count[i];
int cj = count[j];
int q = nz_start[i];
int k;
for (k = 0; k < ci; k++)
if (nonzero[si + k])
model.SupportVectorCoefficients[j - 1][q++] = f[p].alpha[k];
q = nz_start[j];
for (k = 0; k < cj; k++)
if (nonzero[sj + k])
model.SupportVectorCoefficients[i][q++] = f[p].alpha[ci + k];
++p;
}
}
return model;
}
public void startSurfTrain()
{
List<FileInfo> trainingFiles = new List<FileInfo>(1000);
DirectoryInfo di = new DirectoryInfo(Constants.base_folder + "train_" + Constants.CIRCLE_TRIANGLE);
DirectoryInfo[] dirs = di.GetDirectories("*");
foreach (DirectoryInfo dir in dirs)
{
int i = 0;
FileInfo[] files = dir.GetFiles("*.bmp");
foreach (FileInfo fi in files)
{
trainingFiles.Add(fi);
if (i++ > Constants.MAX_TRAIN_SAMPLE)
break;
}
}
double[] class_labels = new double[trainingFiles.Count];
Node[][] nodes = new Node[trainingFiles.Count][];
for (int i = 0; i < trainingFiles.Count; i++)
{
Bitmap bmp = (Bitmap)Bitmap.FromFile(trainingFiles[i].FullName, false);
int com_x_sum = 0, com_y_sum = 0, com_x_y_point_count = 0;
System.Drawing.Imaging.BitmapData image_data = bmp.LockBits(new Rectangle(0, 0, bmp.Width, bmp.Height), System.Drawing.Imaging.ImageLockMode.ReadWrite, bmp.PixelFormat);
int bpp = 3;
int nOffset = image_data.Stride - bmp.Width * bpp;
System.IntPtr Scan0 = image_data.Scan0;
unsafe
{
byte* p = (byte*)Scan0;
for (int y = 0; y < Constants.SIGN_HEIGHT; y++)
{
for (int x = 0; x < Constants.SIGN_WIDTH; x++, p += bpp)
{
if (p[2] == 0)
{
com_x_sum += x;
com_y_sum += y;
com_x_y_point_count++;
}
}
p += nOffset;
}
}
bmp.UnlockBits(image_data);
int com_x = com_x_sum / com_x_y_point_count;
int com_y = com_y_sum / com_x_y_point_count;
Node[] nds = new Node[NNTrain.numOfinputs];
nodes[i] = nds;
bmp.Tag = trainingFiles[i].Name;
fillFeatures_SURF(bmp, com_x, com_y, nds);
class_labels[i] = Double.Parse(trainingFiles[i].Directory.Name);
}
Problem problem = new Problem(nodes.Length, class_labels, nodes, NNTrain.numOfinputs + 1);
// RangeTransform range = Scaling.DetermineRange(problem);
// problem = Scaling.Scale(problem, range);
Parameter param = new Parameter();
param.KernelType = KernelType.POLY;
// param.KernelType = KernelType.LINEAR;
// param.KernelType = KernelType.RBF;
param.SvmType = SvmType.NU_SVC;
param.C = 2;
param.Gamma = .5;
//param.KernelType = KernelType.POLY;
/* double C, Gamma;
ParameterSelection.Grid(problem, param, Constants.base_folder + "params_" + type + ".txt", out C, out Gamma);
param.C = C;
param.Gamma = Gamma;
//param.Probability = true;
*/
Model model = Training.Train(problem, param);
Stream stream = new FileStream(Constants.base_folder + Constants.NN_SVM_SURF + "_" + Constants.CIRCLE_TRIANGLE + ".dat", FileMode.Create, FileAccess.Write, FileShare.None);
BinaryFormatter b = new BinaryFormatter();
b.Serialize(stream, model);
stream.Close();
}
///
public override void Train()
{
int num_users = Feedback.UserMatrix.NumberOfRows; // DH: should be based on MaxUserID for cold case? TODO: investigate.
int num_items = Feedback.ItemMatrix.NumberOfRows;
var svm_features = new List<Node[]>();
Node[][] svm_features_array = svm_features.ToArray();
var svm_parameters = new Parameter();
svm_parameters.SvmType = SvmType.EPSILON_SVR;
//svm_parameters.SvmType = SvmType.NU_SVR;
svm_parameters.C = this.c;
svm_parameters.Gamma = this.gamma;
// user-wise training
this.models = new Model[num_users];
for (int u = 0; u < num_users; u++)
{
var targets = new double[num_items];
for (int i = 0; i < num_items; i++)
targets[i] = Feedback.UserMatrix[u, i] ? 1 : 0;
Problem svm_problem = new Problem(svm_features.Count, targets, svm_features_array, NumItemAttributes - 1); // TODO check
models[u] = SVM.Training.Train(svm_problem, svm_parameters);
}
}
/// <summary>
/// Performs a Grid parameter selection, trying all possible combinations of the two lists and returning the
/// combination which performed best. Use this method if there is no validation data available, and it will
/// divide it 5 times to allow 5-fold validation (training on 4/5 and validating on 1/5, 5 times).
/// </summary>
/// <param name="problem">The training data</param>
/// <param name="parameters">The parameters to use when optimizing</param>
/// <param name="CValues">The set of C values to use</param>
/// <param name="GammaValues">The set of Gamma values to use</param>
/// <param name="outputFile">Output file for the parameter results.</param>
/// <param name="C">The optimal C value will be put into this variable</param>
/// <param name="Gamma">The optimal Gamma value will be put into this variable</param>
public static void Grid(
Problem problem,
Parameter parameters,
List<double> CValues,
List<double> GammaValues,
string outputFile,
out double C,
out double Gamma)
{
Grid(problem, parameters, CValues, GammaValues, outputFile, NFOLD, out C, out Gamma);
}
private void backgroundWorker_DoWork(object sender, DoWorkEventArgs e)
{
Problem problem = new Problem(_X.Count, _Y.ToArray(), _X.ToArray(), 2);
RangeTransform range = RangeTransform.Compute(problem);
problem = range.Scale(problem);
Parameter param = new Parameter();
param.C = 2;
param.Gamma = .5;
Model model = Training.Train(problem, param);
Model.Write("model.txt", model);
int rows = ClientSize.Height;
int columns = ClientSize.Width;
Bitmap image = new Bitmap(columns, rows);
int centerR = rows / 2;
int centerC = columns / 2;
BitmapData buf = image.LockBits(new Rectangle(0, 0, columns, rows), ImageLockMode.WriteOnly, PixelFormat.Format24bppRgb);
unsafe
{
byte* ptr = (byte*)buf.Scan0;
int stride = buf.Stride;
for (int r = 0; r < rows; r++)
{
byte* scan = ptr;
for (int c = 0; c < columns; c++)
{
int x = c - centerC;
int y = r - centerR;
Node[] test = new Node[] { new Node(1, x), new Node(2, y) };
test = range.Transform(test);
int assignment = (int)Prediction.Predict(model, test);
//int assignment = (int)Prediction.Predict(problem, "predict.txt", model, test);
*scan++ = CLASS_FILL[assignment].B;
*scan++ = CLASS_FILL[assignment].G;
*scan++ = CLASS_FILL[assignment].R;
}
ptr += stride;
}
}
image.UnlockBits(buf);
lock (this)
{
_canvas = new Bitmap(image);
}
}
/// <summary>
/// Performs a Grid parameter selection, trying all possible combinations of the two lists and returning the
/// combination which performed best.
/// </summary>
/// <param name="problem">The training data</param>
/// <param name="validation">The validation data</param>
/// <param name="parameters">The parameters to use when optimizing</param>
/// <param name="CValues">The C values to use</param>
/// <param name="GammaValues">The Gamma values to use</param>
/// <param name="outputFile">The output file for the parameter results</param>
/// <param name="C">The optimal C value will be placed in this variable</param>
/// <param name="Gamma">The optimal Gamma value will be placed in this variable</param>
public static void Grid(
Problem problem,
Problem validation,
Parameter parameters,
List<double> CValues,
List<double> GammaValues,
string outputFile,
out double C,
out double Gamma)
{
C = 0;
Gamma = 0;
double maxScore = double.MinValue;
StreamWriter output = null;
if(outputFile != null)
output = new StreamWriter(outputFile);
for (int i = 0; i < CValues.Count; i++)
for (int j = 0; j < GammaValues.Count; j++)
{
parameters.C = CValues[i];
parameters.Gamma = GammaValues[j];
Model model = Training.Train(problem, parameters);
double test = Prediction.Predict(validation, "tmp.txt", model, false);
Console.Write("{0} {1} {2}", parameters.C, parameters.Gamma, test);
if(output != null)
output.WriteLine("{0} {1} {2}", parameters.C, parameters.Gamma, test);
if (test > maxScore)
{
C = parameters.C;
Gamma = parameters.Gamma;
maxScore = test;
Console.WriteLine(" New Maximum!");
}
else Console.WriteLine();
}
if(output != null)
output.Close();
}
public SVC_Q(Problem prob, Parameter param, sbyte[] y_)
: base(prob.Count, prob.X, param)
{
y = (sbyte[])y_.Clone();
cache = new Cache(prob.Count, (long)(param.CacheSize * (1 << 20)));
QD = new float[prob.Count];
for (int i = 0; i < prob.Count; i++)
QD[i] = (float)KernelFunction(i, i);
}
public SVR_Q(Problem prob, Parameter param)
: base(prob.Count, prob.X, param)
{
l = prob.Count;
cache = new Cache(l, (long)(param.CacheSize * (1 << 20)));
QD = new float[2 * l];
sign = new sbyte[2 * l];
index = new int[2 * l];
for (int k = 0; k < l; k++)
{
sign[k] = 1;
sign[k + l] = -1;
index[k] = k;
index[k + l] = k;
QD[k] = (float)KernelFunction(k, k);
QD[k + l] = QD[k];
}
buffer = new float[2][];
buffer[0] = new float[2 * l];
buffer[1] = new float[2 * l];
next_buffer = 0;
}
private static void solve_one_class(Problem prob, Parameter param,
double[] alpha, Solver.SolutionInfo si)
{
int l = prob.Count;
double[] zeros = new double[l];
sbyte[] ones = new sbyte[l];
int i;
int n = (int)(param.Nu * prob.Count); // # of alpha's at upper bound
for (i = 0; i < n; i++)
alpha[i] = 1;
if (n < prob.Count)
alpha[n] = param.Nu * prob.Count - n;
for (i = n + 1; i < l; i++)
alpha[i] = 0;
for (i = 0; i < l; i++)
{
zeros[i] = 0;
ones[i] = 1;
}
Solver s = new Solver();
s.Solve(l, new ONE_CLASS_Q(prob, param), zeros, ones, alpha, 1.0, 1.0, param.EPS, si, param.Shrinking);
}
public static string svm_check_parameter(Problem prob, Parameter param)
{
// svm_type
SvmType svm_type = param.SvmType;
// kernel_type, degree
//KernelType kernel_type = param.KernelType;
if (param.Degree < 0)
return "degree of polynomial kernel < 0";
// cache_size,eps,C,nu,p,shrinking
if (param.CacheSize <= 0)
return "cache_size <= 0";
if (param.EPS <= 0)
return "eps <= 0";
if (param.Gamma == 0)
param.Gamma = 1.0 / prob.MaxIndex;
if (svm_type == SvmType.C_SVC ||
svm_type == SvmType.EPSILON_SVR ||
svm_type == SvmType.NU_SVR)
if (param.C <= 0)
return "C <= 0";
if (svm_type == SvmType.NU_SVC ||
svm_type == SvmType.ONE_CLASS ||
svm_type == SvmType.NU_SVR)
if (param.Nu <= 0 || param.Nu > 1)
return "nu <= 0 or nu > 1";
if (svm_type == SvmType.EPSILON_SVR)
if (param.P < 0)
return "p < 0";
if (param.Probability &&
svm_type == SvmType.ONE_CLASS)
return "one-class SVM probability output not supported yet";
// check whether nu-svc is feasible
if (svm_type == SvmType.NU_SVC)
{
int l = prob.Count;
int Max_nr_class = 16;
int nr_class = 0;
int[] label = new int[Max_nr_class];
int[] count = new int[Max_nr_class];
int i;
for (i = 0; i < l; i++)
{
int this_label = (int)prob.Y[i];
int j;
for (j = 0; j < nr_class; j++)
if (this_label == label[j])
{
++count[j];
break;
}
if (j == nr_class)
{
if (nr_class == Max_nr_class)
{
Max_nr_class *= 2;
int[] new_data = new int[Max_nr_class];
Array.Copy(label, 0, new_data, 0, label.Length);
label = new_data;
new_data = new int[Max_nr_class];
Array.Copy(count, 0, new_data, 0, count.Length);
count = new_data;
}
label[nr_class] = this_label;
count[nr_class] = 1;
++nr_class;
}
}
for (i = 0; i < nr_class; i++)
{
int n1 = count[i];
for (int j = i + 1; j < nr_class; j++)
{
int n2 = count[j];
if (param.Nu * (n1 + n2) / 2 > Math.Min(n1, n2))
return "specified nu is infeasible";
}
}
}
return null;
}
// Cross-validation decision values for probability estimates
private static void svm_binary_svc_probability(Problem prob, Parameter param, double Cp, double Cn, double[] probAB)
{
int i;
int nr_fold = 5;
int[] perm = new int[prob.Count];
double[] dec_values = new double[prob.Count];
// random shuffle
Random rand = new Random();
for (i = 0; i < prob.Count; i++) perm[i] = i;
for (i = 0; i < prob.Count; i++)
{
int j = i + (int)(rand.NextDouble() * (prob.Count - i));
do { int _ = perm[i]; perm[i] = perm[j]; perm[j] = _; } while (false);
}
for (i = 0; i < nr_fold; i++)
{
int begin = i * prob.Count / nr_fold;
int end = (i + 1) * prob.Count / nr_fold;
int j, k;
Problem subprob = new Problem();
subprob.Count = prob.Count - (end - begin);
subprob.X = new Node[subprob.Count][];
subprob.Y = new double[subprob.Count];
k = 0;
for (j = 0; j < begin; j++)
{
subprob.X[k] = prob.X[perm[j]];
subprob.Y[k] = prob.Y[perm[j]];
++k;
}
for (j = end; j < prob.Count; j++)
{
subprob.X[k] = prob.X[perm[j]];
subprob.Y[k] = prob.Y[perm[j]];
++k;
}
int p_count = 0, n_count = 0;
for (j = 0; j < k; j++)
if (subprob.Y[j] > 0)
p_count++;
else
n_count++;
if (p_count == 0 && n_count == 0)
for (j = begin; j < end; j++)
dec_values[perm[j]] = 0;
else if (p_count > 0 && n_count == 0)
for (j = begin; j < end; j++)
dec_values[perm[j]] = 1;
else if (p_count == 0 && n_count > 0)
for (j = begin; j < end; j++)
dec_values[perm[j]] = -1;
else
{
Parameter subparam = (Parameter)param.Clone();
subparam.Probability = false;
subparam.C = 1.0;
subparam.Weights[1] = Cp;
subparam.Weights[-1] = Cn;
Model submodel = svm_train(subprob, subparam);
for (j = begin; j < end; j++)
{
double[] dec_value = new double[1];
svm_predict_values(submodel, prob.X[perm[j]], dec_value);
dec_values[perm[j]] = dec_value[0];
// ensure +1 -1 order; reason not using CV subroutine
dec_values[perm[j]] *= submodel.ClassLabels[0];
}
}
}
sigmoid_train(prob.Count, dec_values, prob.Y, probAB);
}
private static void solve_epsilon_svr(Problem prob, Parameter param, double[] alpha, Solver.SolutionInfo si)
{
int l = prob.Count;
double[] alpha2 = new double[2 * l];
double[] linear_term = new double[2 * l];
sbyte[] y = new sbyte[2 * l];
int i;
for (i = 0; i < l; i++)
{
alpha2[i] = 0;
linear_term[i] = param.P - prob.Y[i];
y[i] = 1;
alpha2[i + l] = 0;
linear_term[i + l] = param.P + prob.Y[i];
y[i + l] = -1;
}
Solver s = new Solver();
s.Solve(2 * l, new SVR_Q(prob, param), linear_term, y, alpha2, param.C, param.C, param.EPS, si, param.Shrinking);
double sum_alpha = 0;
for (i = 0; i < l; i++)
{
alpha[i] = alpha2[i] - alpha2[i + l];
sum_alpha += Math.Abs(alpha[i]);
}
Procedures.info("nu = " + sum_alpha / (param.C * l) + "\n");
}
// Return parameter of a Laplace distribution
private static double svm_svr_probability(Problem prob, Parameter param)
{
int i;
int nr_fold = 5;
double[] ymv = new double[prob.Count];
double mae = 0;
Parameter newparam = (Parameter)param.Clone();
newparam.Probability = false;
svm_cross_validation(prob, newparam, nr_fold, ymv);
for (i = 0; i < prob.Count; i++)
{
ymv[i] = prob.Y[i] - ymv[i];
mae += Math.Abs(ymv[i]);
}
mae /= prob.Count;
double std = Math.Sqrt(2 * mae * mae);
int count = 0;
mae = 0;
for (i = 0; i < prob.Count; i++)
if (Math.Abs(ymv[i]) > 5 * std)
count = count + 1;
else
mae += Math.Abs(ymv[i]);
mae /= (prob.Count - count);
Procedures.info("Prob. model for test data: target value = predicted value + z,\nz: Laplace distribution e^(-|z|/sigma)/(2sigma),sigma=" + mae + "\n");
return mae;
}
private static void solve_nu_svr(Problem prob, Parameter param,
double[] alpha, Solver.SolutionInfo si)
{
int l = prob.Count;
double C = param.C;
double[] alpha2 = new double[2 * l];
double[] linear_term = new double[2 * l];
sbyte[] y = new sbyte[2 * l];
int i;
double sum = C * param.Nu * l / 2;
for (i = 0; i < l; i++)
{
alpha2[i] = alpha2[i + l] = Math.Min(sum, C);
sum -= alpha2[i];
linear_term[i] = -prob.Y[i];
y[i] = 1;
linear_term[i + l] = prob.Y[i];
y[i + l] = -1;
}
Solver_NU s = new Solver_NU();
s.Solve(2 * l, new SVR_Q(prob, param), linear_term, y, alpha2, C, C, param.EPS, si, param.Shrinking);
Procedures.info("epsilon = " + (-si.r) + "\n");
for (i = 0; i < l; i++)
alpha[i] = alpha2[i] - alpha2[i + l];
}
/// <summary>
/// Performs a Grid parameter selection, trying all possible combinations of the two lists and returning the
/// combination which performed best. Use this method if validation data isn't available, as it will
/// divide the training data and train on a portion of it and test on the rest.
/// </summary>
/// <param name="problem">The training data</param>
/// <param name="parameters">The parameters to use when optimizing</param>
/// <param name="CValues">The set of C values to use</param>
/// <param name="GammaValues">The set of Gamma values to use</param>
/// <param name="outputFile">Output file for the parameter results.</param>
/// <param name="nrfold">The number of times the data should be divided for validation</param>
/// <param name="C">The optimal C value will be placed in this variable</param>
/// <param name="Gamma">The optimal Gamma value will be placed in this variable</param>
public static void Grid(
Problem problem,
Parameter parameters,
List<double> CValues,
List<double> GammaValues,
string outputFile,
int nrfold,
out double C,
out double Gamma)
{
C = 0;
Gamma = 0;
double crossValidation = double.MinValue;
StreamWriter output = null;
if(outputFile != null)
output = new StreamWriter(outputFile);
for(int i=0; i<CValues.Count; i++)
for (int j = 0; j < GammaValues.Count; j++)
{
parameters.C = CValues[i];
parameters.Gamma = GammaValues[j];
double test = Training.PerformCrossValidation(problem, parameters, nrfold);
if(output != null)
output.WriteLine("{0} {1} {2}", parameters.C, parameters.Gamma, test);
if (test > crossValidation)
{
C = parameters.C;
Gamma = parameters.Gamma;
crossValidation = test;
}
}
if(output != null)
output.Close();
}