/// <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);
}
/// <summary>
/// Determines the Gaussian transform for the provided problem.
/// </summary>
/// <param name="prob">The Problem to analyze</param>
/// <returns>The Gaussian transform for the problem</returns>
public static GaussianTransform Compute(Problem prob)
{
int[] counts = new int[prob.MaxIndex];
double[] means = new double[prob.MaxIndex];
foreach (Node[] sample in prob.X) {
for (int i = 0; i < sample.Length; i++) {
means[sample[i].Index - 1] += sample[i].Value;
counts[sample[i].Index - 1]++;
}
}
for (int i = 0; i < prob.MaxIndex; i++) {
if (counts[i] == 0)
counts[i] = 2;
means[i] /= counts[i];
}
double[] stddevs = new double[prob.MaxIndex];
foreach (Node[] sample in prob.X) {
for (int i = 0; i < sample.Length; i++) {
double diff = sample[i].Value - means[sample[i].Index - 1];
stddevs[sample[i].Index - 1] += diff * diff;
}
}
for (int i = 0; i < prob.MaxIndex; i++) {
if (stddevs[i] == 0)
continue;
stddevs[i] /= (counts[i] - 1);
stddevs[i] = Math.Sqrt(stddevs[i]);
}
return new GaussianTransform(means, stddevs);
}
/// <summary>
/// Scales a problem using the provided range. This will not affect the parameter.
/// </summary>
/// <param name="prob">The problem to scale</param>
/// <param name="range">The Range transform to use in scaling</param>
/// <returns>The Scaled problem</returns>
public static Problem Scale(this IRangeTransform range, Problem prob)
{
Problem scaledProblem = new Problem(prob.Count, new double[prob.Count], new Node[prob.Count][], prob.MaxIndex);
for (int i = 0; i < scaledProblem.Count; i++)
{
scaledProblem.X[i] = new Node[prob.X[i].Length];
for (int j = 0; j < scaledProblem.X[i].Length; j++)
scaledProblem.X[i][j] = new Node(prob.X[i][j].Index, range.Transform(prob.X[i][j].Value, prob.X[i][j].Index));
scaledProblem.Y[i] = prob.Y[i];
}
return scaledProblem;
}
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;
}
// 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);
}
/// <summary>
/// Predicts the class memberships of all the vectors in the problem.
/// </summary>
/// <param name="problem">The SVM Problem to solve</param>
/// <param name="outputFile">File for result output</param>
/// <param name="model">The Model to use</param>
/// <param name="predict_probability">Whether to output a distribution over the classes</param>
/// <returns>Percentage correctly labelled</returns>
public static double Predict(
Problem problem,
string outputFile,
Model model,
bool predict_probability)
{
int correct = 0;
int total = 0;
double error = 0;
double sumv = 0, sumy = 0, sumvv = 0, sumyy = 0, sumvy = 0;
StreamWriter output = outputFile != null ? new StreamWriter(outputFile) : null;
SvmType svm_type = Procedures.svm_get_svm_type(model);
int nr_class = Procedures.svm_get_nr_class(model);
int[] labels = new int[nr_class];
double[] prob_estimates = null;
if (predict_probability)
{
if (svm_type == SvmType.EPSILON_SVR || svm_type == SvmType.NU_SVR)
{
Console.WriteLine("Prob. model for test data: target value = predicted value + z,\nz: Laplace distribution e^(-|z|/sigma)/(2sigma),sigma=" + Procedures.svm_get_svr_probability(model));
}
else
{
Procedures.svm_get_labels(model, labels);
prob_estimates = new double[nr_class];
if (output != null)
{
output.Write("labels");
for (int j = 0; j < nr_class; j++)
{
output.Write(" " + labels[j]);
}
output.Write("\n");
}
}
}
for (int i = 0; i < problem.Count; i++)
{
double target = problem.Y[i];
Node[] x = problem.X[i];
double v;
if (predict_probability && (svm_type == SvmType.C_SVC || svm_type == SvmType.NU_SVC))
{
v = Procedures.svm_predict_probability(model, x, prob_estimates);
if (output != null)
{
output.Write(v + " ");
for (int j = 0; j < nr_class; j++)
{
output.Write(prob_estimates[j] + " ");
}
output.Write("\n");
}
}
else
{
v = Procedures.svm_predict(model, x);
if(output != null)
output.Write(v + "\n");
}
if (v == target)
++correct;
error += (v - target) * (v - target);
sumv += v;
sumy += target;
sumvv += v * v;
sumyy += target * target;
sumvy += v * target;
++total;
}
if(output != null)
output.Close();
return (double)correct / total;
}
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.
/// </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 Problem CuckooSearch(Problem prob, out double storagePercentage)
{
int nNests = 5; //number of nests, or number of solutions
int subsetSize = 100;
int maxGen = 5; //maximum generation
double discoveryRate = 0.25; //discovery rate of alien eggs
double tolerance = Math.Exp(-5);
int lowerBound = -5;
int upperBound = 5;
int totalInstances = prob.X.Count(); //problem size
double[] cuckooFitnessVal = new double[nNests];
double[] newCuckooFitnessVal = new double[nNests];
ObjectInstanceSelection globalBestCuckoo = null;
double globalBest = double.MinValue;
Random rand = new Random();
FlowerPollinationAlgorithm fpa = new FlowerPollinationAlgorithm();
//initialize population
List <ObjectInstanceSelection> cuckoos = InitializeBinaryCuckoo(nNests, subsetSize, totalInstances, prob);
List <ObjectInstanceSelection> newCuckoos = new List <ObjectInstanceSelection>(cuckoos.Count); //create a clone of bats
cuckoos.ForEach((item) =>
{
newCuckoos.Add(new ObjectInstanceSelection(item.Attribute_Values, item.Attribute_Values_Continuous, item.Pointers, item.Fitness)); //create a clone of flowers
});
cuckooFitnessVal = EvaluateObjectiveFunction(cuckoos, prob); //evaluate fitness value for all the bats
newCuckooFitnessVal = EvaluateObjectiveFunction(newCuckoos, prob); //evaluate fitness value for new bats. Note: this will be the same for this function call, since pollination has not occur
CuckooFitness(cuckooFitnessVal, cuckoos); //fitness value for each bats
CuckooFitness(newCuckooFitnessVal, newCuckoos); //fitness value for new bats
globalBestCuckoo = EvaluateSolution(cuckooFitnessVal, newCuckooFitnessVal, globalBest, cuckoos, newCuckoos, globalBestCuckoo); //get the global best flower
globalBest = globalBestCuckoo.__Fitness;
//generate new solutions
double beta = 3 / 2;
double A = fp.Gamma(1 + beta) * Math.Sin(Math.PI * (beta / 2));
double B = fp.Gamma((1 + beta) / 2) * beta;
double C = (beta - 1) / 2;
double D = Math.Pow(2, C);
double E = A / (B * D);
double sigma = Math.Pow(E, (1 / beta));
double F;
double G;
double step;
double stepSize;
int x = 0;
for (int i = 0; i <= maxGen; i++)
{
for (int j = 0; j < nNests; j++)
{
for (int k = 0; k < subsetSize; k++)
{
F = SimpleRNG.GetNormal() * sigma;
G = SimpleRNG.GetNormal();
step = F / Math.Pow(Math.Abs(G), (1 / beta));
//In the next equation, the difference factor (s-best) means that when the solution is the best solution, it remains unchanged.
//Here the factor 0.01 comes from the fact that L/100 should the typical step size of walks/flights where L is the typical lenghtscale;
//otherwise, Levy flights may become too aggresive/efficient, which makes new solutions (even) jump out side of the design domain (and thus wasting evaluations).
stepSize = 0.01 * step * (cuckoos[j].Attribute_Values[k] - globalBestCuckoo.Attribute_Values[k]);
//Now the actual random walks or levyy flights
newCuckoos[j].Attribute_Values[k] = fi.Binarize((newCuckoos[j].Attribute_Values[k] + stepSize) * SimpleRNG.GetNormal(), rand.NextDouble());
if (cuckoos[j].Attribute_Values[k] == 1 && newCuckoos[j].Attribute_Values[k] == 0)
{
x++;
}
}
}
//discovery and randomization - replace some nest by constructing new solutions
newCuckoos = EmptyNest(cuckoos, newCuckoos, discoveryRate, subsetSize, nNests);
//Select best solutions from the original population and matured population for the next generation;
fpa.SelectBestSolution(cuckoos, newCuckoos);
//evaluate new solution
newCuckooFitnessVal = EvaluateObjectiveFunction(newCuckoos, prob); //evaluate fitness value for all the bats
CuckooFitness(newCuckooFitnessVal, newCuckoos); //fitness value for new bats
globalBestCuckoo = EvaluateSolution(cuckooFitnessVal, newCuckooFitnessVal, globalBest, cuckoos, newCuckoos, globalBestCuckoo); //get the global best flower
globalBest = globalBestCuckoo.Fitness;
//if solution has converged to a optimal user-defined point, stop search
int Max = 60; // maximum percentage reduction
if (globalBest >= Max) //if the percentage reduction has approached 60%, stop search!
{
break;
}
}
//ensure that at least, N instances are selected for classification
int min = 40; //minimum number of selected instances
globalBestCuckoo = fpa.AddInstances(globalBestCuckoo, min);
Problem subBest = fi.buildModelMultiClass(globalBestCuckoo, prob); //build model for the best Instance Mast
storagePercentage = Training.StoragePercentage(subBest, prob); //calculate the percent of the original training set was retained by the reduction algorithm
return(subBest);
}
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];
}
//compute the k-nearest neighbour of all instances in the dataset
public Problem computeNearestNeighbour(int k, Problem trainDataset, int numOfSubset)
{
double sum = 0;
double distance;
int n = trainDataset.Count; //number of data instances
int subN = 400;
//int n = 800; //number of data instances
List <Node[]> nearestNeighbours = new List <Node[]>();
List <double> dist = new List <double>();
List <double> labels = new List <double>();
List <int> index = new List <int>();
Node[] xNodes = new Node[n];
Node[] yNodes = new Node[n];
object[,] obj = new object[subN, 4];
List <object[, ]> objList = new List <object[, ]>();
object[,] temp = new object[1, 3];
List <Problem> ds = new List <Problem>();
object[,] nn = new object[n, 7]; //data structure containing the NNs and their corresponding distances
double score = 0; //score assigned to individual instance by the oppositiley NNs in its neighbourhood list
object[,] scoreList = new object[n, 4]; //scores assigned to all the instances
object[,] dataSubset = new object[n, 3]; //subset of data to return
//Get the neighbourhood list of all the instances in the dataset. That is, compute distance between Xi and other instances in the dataset.
List <object[, ]> objSortedList = new List <object[, ]>();
for (int i = 0; i < n; i++)
{
int ctr = 0; int cntr1 = 0; int cntr2 = 0; int a = 0; int b = 0;
int countP = trainDataset.Y.Count(q => q == 1);
int countN = trainDataset.Y.Count(q => q == -1);
//generate unique random number
//List<int> rNum = Training.GetRandomNumbers(2000, n);
for (int j = 0; j < n; j++)
{
if (j.Equals(i))
{
continue;
}
//randomly select N instances from dataset
else if (cntr1 < (subN * 0.5) && trainDataset.Y[j] == 1) //compute distance for positive class (50% of k goes for positive instances)
{
distance = Kernel.computeSquaredDistance(trainDataset.X[i], trainDataset.X[j]); //compute the distance between Xi and all other instances in the dataset
obj[ctr, 0] = distance;
obj[ctr, 1] = trainDataset.X[j];
obj[ctr, 2] = trainDataset.Y[j];
obj[ctr, 3] = ctr; //save the index
ctr++; //save the instance and their corresponding distances
cntr1++;
}
else if (cntr2 < (subN * 0.5) && trainDataset.Y[j] == -1) //compute distance for negative class (50% of k goes for negative instances)
{
distance = Kernel.computeSquaredDistance(trainDataset.X[i], trainDataset.X[j]); //compute the distance between Xi and all other instances in the dataset
obj[ctr, 0] = distance;
obj[ctr, 1] = trainDataset.X[j];
obj[ctr, 2] = trainDataset.Y[j];
obj[ctr, 3] = ctr; //save the index
ctr++; //save the instance and their corresponding distances
cntr2++;
}
//distance = Kernel.computeSquaredDistance(trainDataset.X[i], trainDataset.X[j]); //compute the distance between Xi and all other instances in the dataset
////save the instance and their corresponding distances
//obj[a, 0] = distance;
//obj[a, 1] = trainDataset.X[j];
//obj[a, 2] = trainDataset.Y[j];
//obj[a, 3] = a; //save the index
//a++;
}
objList.Add(obj); //Data structure (or List), containing the instances and distances of K nearest neighbours of every instance in the dataset
obj = new object[subN, 4];
}
//sort the data structure. That, sort(and retain index) the neighbourhood list of each instance in the dataset
for (int i = 0; i < subN; i++)
{
object[,] objSort = sortMultiArray(objList[i]); //sort array to select the nearest neighbour of Xi
objSortedList.Add(objSort); //add to list
}
//select boundary instances
for (int i = 0; i < n; i++)
{
//select the k-neareast neighbours (using top K elements), their corresponding distances and class labels of Xi
int subK = k;
int count1 = 0; int count2 = 0;
for (int p = 0; p < subN; p++)
{
object[,] objSorted = objSortedList[p];
if (count1 < (subK / 2) && (double)objSorted[p, 2] == 1) //50% of k goes to positive class. This is to ensure that there is a balance in the training subset
{
dist.Add((double)objSorted[p, 0]); //distance
nearestNeighbours.Add((Node[])objSorted[p, 1]); //nearest neighbour i
labels.Add((double)objSorted[p, 2]); //class labels
index.Add((int)objSorted[p, 3]); //add index for each nearest neighbour
count1++;
}
else if (count2 < (subK / 2) && (double)objSorted[p, 2] == -1) //50% of K goes to negative class
{
dist.Add((double)objSorted[p, 0]); //distance
nearestNeighbours.Add((Node[])objSorted[p, 1]); //nearest neighbour i
labels.Add((double)objSorted[p, 2]); //class labels
index.Add((int)objSorted[p, 3]); //add index for each nearest neighbour
count2++;
}
}
nn[i, 0] = k;
nn[i, 1] = dist;
nn[i, 2] = nearestNeighbours;
nn[i, 3] = trainDataset.X[i];
nn[i, 4] = labels;
nn[i, 5] = trainDataset.Y[i];
nn[i, 6] = index; //save the index
//Compute Exponential Decay
double EDScore = 0; //Exponential decay score
int counter = 0;
double distNN = 0;
List <double> distNNList = new List <double>();
for (int p = 0; p < subK; p++)
{
//compute exponential decay for Xi and all its Nearest neighbour belonging to the opposite class
//if the label of the current instance in the neighbourhood list is not equal to the label of ith instance then compute its Exponential Decay Score
if (((List <double>)nn[i, 4])[p] != (double)nn[i, 5]) //identify the nearest neighbour belonging to the opposite class
{
int indx = ((List <int>)nn[i, 6])[p]; //get the index of the current nearest neighbour
object[,] objNN = objSortedList[indx]; //get the current nearest neighbour from list
//using the index, select the distance of the closest instance of the opposite class on its neighborhood list
for (int a = 0; a < subN; a++)
{
double label1 = (double)objNN[a, 2]; //label of the current instance in the neighbourhood list of the current nearest neigbour
double label2 = ((List <double>)nn[i, 4])[p]; //label of the current instance in the neighbourhood list
//if the statement below is true (that is, if the labels are not equal), then select the closest instance of the opposite class on its neighborhood list
//List is ordered already, hence the topmost instance of the opposite class in the neighbourhood list, is the closest instance
if (label1 != label2)
{
distNN = (double)objNN[a, 0]; //get the distance and break. We only need the distance of the closest instance.
distNNList.Add(distNN);
break;
}
}
EDScore += ((List <double>)nn[i, 1])[p] - Math.Pow(distNN, 2); //compute exponential decay score
//EDScore += ((List<double>)nn[i, 1])[p] - Math.Pow(((List<double>)nn[i, 1])[p], 2); //compute exponential decay score
counter++;
}
}
EDScore = EDScore / counter;
//determine the scores of every instance
int numOfContributors = counter; int b = 0;
for (int p = 0; p < subK; p++)
{
//if the label of the current instance in the neighbourhood list is not equal to the label of ith instance
if (((List <double>)nn[i, 4])[p] != (double)nn[i, 5])//identify the nearest neighbour belonging to the opposite class
{
//score += Math.Exp(-(((List<double>)nn[i, 1])[p] - Math.Pow(((List<double>)nn[i, 1])[p], 2) / EDScore));
score += Math.Exp(-(((List <double>)nn[i, 1])[p] - Math.Pow(distNNList[b++], 2) / EDScore));
}
}
score = score / numOfContributors;
scoreList[i, 0] = score; scoreList[i, 1] = nn[i, 3]; scoreList[i, 2] = nn[i, 5]; scoreList[i, 3] = nn[i, 6];
dist = new List <double>(); nearestNeighbours = new List <Node[]>(); labels = new List <double>();
}
sortMultiArray(scoreList); //sort scores to select the best N instances to be used for training
//select data subset to be used for training. Selected subset are instances that are closest to the data boundary
Node[][] xScoreList = new Node[numOfSubset][];
double[] yScoreList = new double[numOfSubset];
int cnt1 = 0, cnt2 = 0, cnt3 = 0;
int total = n - 1;
for (int i = 0; i < n; i++)
{
for (int j = 0; j < 3; j++)
{
dataSubset[i, j] = scoreList[total, j]; //select instances with the highest scores
}
if (cnt1 < (0.7 * numOfSubset) && (double)dataSubset[i, 2] == 1) //select 70% positive instance of the subset
{
xScoreList[cnt3] = (Node[])dataSubset[i, 1];
yScoreList[cnt3] = (double)dataSubset[i, 2];
cnt1++; cnt3++;
}
else if (cnt2 < (0.3 * numOfSubset) && (double)dataSubset[i, 2] == -1) //select 30% negative instance of the subset
{
xScoreList[cnt3] = (Node[])dataSubset[i, 1];
yScoreList[cnt3] = (double)dataSubset[i, 2];
cnt2++; cnt3++;
}
total--;
}
Problem subset = new Problem(numOfSubset, yScoreList, xScoreList, xScoreList[0].GetLength(0));
return(subset);
}
//build model for multi class problems
public Problem buildModelMultiClass(ObjectInstanceSelection firefly, Problem prob)
{
int tNI = firefly.Attribute_Values.Count(); //size of each Instance Mask
List <double> y = new List <double>();
List <Node[]> x = new List <Node[]>();
bool pos = false, neg = false;
List <double> classes = getClassLabels(prob.Y); //get the class labels
int nClass = classes.Count; //count the number of classes
int[] classCount = new int[nClass];
//building model for each instance in instance mask in each firefly object
for (int j = 0; j < tNI; j++)
{
if (firefly.__Attribute_Values[j] == 1) //if instance is selected, use for classification
{
int p = firefly.__Pointers[j];
x.Add(prob.X[p]);
y.Add(prob.Y[p]);
for (int i = 0; i < nClass; i++)
{
if (prob.Y[p] == classes[i])
{
classCount[i]++; //count the total number of instances in each class
}
}
}
else
{
continue;
}
}
Node[][] X = new Node[x.Count][];
double[] Y = new double[y.Count];
//ensuring that the subproblem consist of both positive and negative instance
int k = 0;
if (classCount.Sum() == 0) //if the sum is zero, then no instance was selected
{
return(null);
}
else //ensure that instance mask contains at least, one of each class instance
{
for (int a = 0; a < nClass; a++)
{
if (classCount[a] == 0)
{
int m = 0;
for (int i = 0; i < prob.Count; i++) //if no instance in this class, search the whole subproblem and insert one instance in the kth position of subproblem
{
if (prob.Y[i] == classes[a])
{
x[k] = prob.X[i]; //insert negative instance in the first and second position
y[k] = prob.Y[i]; //insert label
k++; m++;
}
if (m == 2)
{
break;
}
}
}
}
}
x.CopyTo(X); //convert from list to double[] array
y.CopyTo(Y);
Problem subProb = new Problem(X.Count(), Y, X, X[0].GetLength(0));
return(subProb);
}
//compute the k-nearest neighbour of all instances in the dataset
public Problem computeNearestNeighbour(int k, Problem trainDataset, int numOfSubset)
{
double sum = 0; double distance;
int n = trainDataset.Count; //number of data instances
List <Node[]> nearestNeighbours = new List <Node[]>(); List <double> dist = new List <double>(); List <double> labels = new List <double>();
Node[] xNodes = new Node[n];
Node[] yNodes = new Node[n];
object[,] obj = new object[n - 1, 3];
//object[,] obj = new object[k, 3];
object[,] temp = new object[1, 3];
List <Problem> ds = new List <Problem>();
object[,] nn = new object[n, 6]; //data structure containing the NNs and their corresponding distances
double score = 0; //score assigned to individual instance by the oppositiley NNs in its neighbourhood list
object[,] scoreList = new object[n, 3]; //scores assigned to all the instances
object[,] dataSubset = new object[n, 3]; //subset of data to return
//compute distance between Xi and other instances
for (int i = 0; i < n; i++)
{
int ctr = 0; int cntr1 = 0; int cntr2 = 0;
int countP = trainDataset.Y.Count(q => q == 1);
int countN = trainDataset.Y.Count(q => q == -1);
for (int j = 0; j < n; j++)
{
if (j.Equals(i))
{
continue;
}
if (countN <= 1) //come here if we have very few selected negative instance in the subproblem
{
double propP = n * 0.9, propN = n * 0.1;
obj = buildObject(ref ctr, ref cntr1, ref cntr2, i, j, obj, trainDataset, propP, propN); //0.9 and 0.1 are proportion of positive and negative instances to be selected
//ctr++; cntr1++; cntr2++;
}
else if (countP <= 1) //come here if we have very few selected positive instance
{
double propP = n * 0.1, propN = n * 0.9;
obj = buildObject(ref ctr, ref cntr1, ref cntr2, i, j, obj, trainDataset, propP, propN); //0.1 and 0.9 are proportion of positive and negative instances to be selected
}
else if (n > trainDataset.Count) //come here of n is more than the total number of selected instances
{
double propP = countP, propN = trainDataset.Count - countP; //in this case, selected instances consist of all the positive instance and a portion of negative instance
obj = buildObject(ref ctr, ref cntr1, ref cntr2, i, j, obj, trainDataset, propP, propN);
//ctr++; cntr1++; cntr2++;
}
else if (countN < (n * 0.7) || countP < (n * 0.3)) //come here if the selected positive or negative instances is less than the defined proportion
{
if (countP < (n * 0.3))
{
double propP = countP, propN = n - countP; //in this case, selected instances consist of all the positive instance and a portion of negative instance
obj = buildObject(ref ctr, ref cntr1, ref cntr2, i, j, obj, trainDataset, propP, propN);
}
else if (countN < (n * 0.7))
{
double propP = n - countN, propN = countN; //in this case, selected instances consist of all the positive instance and a portion of negative instance
obj = buildObject(ref ctr, ref cntr1, ref cntr2, i, j, obj, trainDataset, propP, propN);
}
}
else //come here if we have fairly good distribution of positive and negative instances
{
double propP = n * 0.3, propN = n * 0.7;
obj = buildObject(ref ctr, ref cntr1, ref cntr2, i, j, obj, trainDataset, propP, propN); //0.3 and 0.7 are proportion of positive and negative instances to be selected
}
}
Training.sortMultiArray(obj); //sort array to select the nearest neighbour of Xi
//select the k-neareast neighbours (using top K elements), their corresponding distances and class labels of Xi
//int subK = 30;
int subK = k;
int count1 = 0; int count2 = 0; int sumN = 0, sumP = 0;
for (int z = 0; z < obj.GetLength(0); z++) //count the total number of positive and negative instances in the subproblem
{
if ((double)obj[z, 2] == 1)
{
sumP++;
}
else
{
sumN++;
}
}
for (int p = 0; p < k; p++) //select k-neareast neighbours (using top K elements), their corresponding distances and class labels of Xi
{
if (count1 < sumP && (double)obj[p, 2] == 1) //NN for positive class
{
dist.Add((double)obj[p, 0]); //distance
nearestNeighbours.Add((Node[])obj[p, 1]); //nearest neighbour i
labels.Add((double)obj[p, 2]); //class labels
count1++;
}
else if (count2 < sumN && (double)obj[p, 2] == -1) // NN for negative class
{
dist.Add((double)obj[p, 0]); //distance
nearestNeighbours.Add((Node[])obj[p, 1]); //nearest neighbour i
labels.Add((double)obj[p, 2]); //class labels
count2++;
}
}
//for (int z = 0; z < obj.Length; z++)
nn[i, 0] = k; nn[i, 1] = dist; nn[i, 2] = nearestNeighbours; nn[i, 3] = trainDataset.X[i]; nn[i, 4] = labels; nn[i, 5] = trainDataset.Y[i];
//Compute Exponential Decay
double EDScore = 0; //Exponential decay score
int counter = 0;
for (int p = 0; p < subK; p++)
{
//compute exponential decay for Xi and all its Nearest neighbour belonging to the opposite class
//if the label of the current instance in the neighbourhood list is not equal to the label of ith instance then compute its Exponential Decay Score
if (((List <double>)nn[i, 4])[p] != (double)nn[i, 5]) //identify the nearest neighbour belonging to the opposite class
{
EDScore += ((List <double>)nn[i, 1])[p] - Math.Pow(((List <double>)nn[i, 1])[p], 2); //compute exponential decay score
counter++; //counting the number of contributors
}
}
EDScore = EDScore / counter;
//determine the scores of every instance
//int numOfContributors = k - counter; //number of NN of opposite class that contributes to Xi
int numOfContributors = counter;
for (int p = 0; p < subK; p++)
{
//if the label of the current instance in the neighbourhood list is not equal to the label of ith instance
if (((List <double>)nn[i, 4])[p] != (double)nn[i, 5])//identify the nearest neighbour belonging to the opposite class
{
score += Math.Exp(-(((List <double>)nn[i, 1])[p] - Math.Pow(((List <double>)nn[i, 1])[p], 2) / EDScore));
}
}
score = score / numOfContributors;
scoreList[i, 0] = score; scoreList[i, 1] = nn[i, 3]; scoreList[i, 2] = nn[i, 5];
dist = new List <double>(); nearestNeighbours = new List <Node[]>(); labels = new List <double>();
//EDScoreList.Add(EDScore);//list of Exponential Decay scores
//Problem pp = new Problem(k, dist, nearestNeighbours, trainDataset.X[i], labels);
//ds.Add(pp);
}
Training.sortMultiArray(scoreList); //sort scores to select the best N instances to be used for training
//select data subset to be used for training. Selected subset are instances that are closest to the data boundary
Node[][] xScoreList = new Node[numOfSubset][];
double[] yScoreList = new double[numOfSubset];
int cnt1 = 0, cnt2 = 0, cnt3 = 0;
int total = n - 1;
for (int i = 0; i < n; i++)
{
for (int j = 0; j < 3; j++)
{
dataSubset[i, j] = scoreList[total, j]; //select instances with the highest scores
}
if (cnt1 < (0.1 * numOfSubset) && (double)dataSubset[i, 2] == 1) //select 70% positive instance of the subset
{
xScoreList[cnt3] = (Node[])dataSubset[i, 1];
yScoreList[cnt3] = (double)dataSubset[i, 2];
cnt1++; cnt3++;
}
else if (cnt2 < (0.9 * numOfSubset) && (double)dataSubset[i, 2] == -1) //select 30% negative instance of the subset
{
xScoreList[cnt3] = (Node[])dataSubset[i, 1];
yScoreList[cnt3] = (double)dataSubset[i, 2];
cnt2++; cnt3++;
}
total--;
}
Problem subset = new Problem(numOfSubset, yScoreList, xScoreList, xScoreList[0].GetLength(0));
return(subset);
}
public object[,] buildObject(ref int ctr, ref int cntr1, ref int cntr2, int i, int j, object[,] obj, Problem trainDataset, double propP, double propN)
{
//ctr = a; cntr1 = b; cntr2 = c;
double distance;
if (cntr1 < propP && trainDataset.Y[j] == 1) //compute distance for positive class (90% of k goes for positive instances)
{
distance = Kernel.computeSquaredDistance(trainDataset.X[i], trainDataset.X[j]); //compute the distance between Xi and all other instances in the dataset
obj[ctr, 0] = distance; obj[ctr, 1] = trainDataset.X[j]; obj[ctr, 2] = trainDataset.Y[j]; //save the instance and their corresponding distances
ctr++; cntr1++;
}
else if (trainDataset.Y[j] == -1 && cntr2 < propN) //compute distance for negative class (10% of k goes for negative instances)
{
distance = Kernel.computeSquaredDistance(trainDataset.X[i], trainDataset.X[j]); //compute the distance between Xi and all other instances in the dataset
obj[ctr, 0] = distance; obj[ctr, 1] = trainDataset.X[j]; obj[ctr, 2] = trainDataset.Y[j]; //save the instance and their corresponding distances
ctr++; cntr2++;
}
return(obj);
}
/// <summary>
/// Move all fireflies toward brighter ones
/// </summary>
//public void ffa_move(double[] Lightn, ObjectInstanceSelection[] fireflies0, double[] Lighto, double alpha, double gamma, List<ObjectInstanceSelection> fireflies,
// Problem prob, Parameter param, List<double> avgAcc, List<int> changedIndex)
public void ffa_move(double[] Lightn, ObjectInstanceSelection[] fireflies0, double[] Lighto, double alpha, double gamma, List <ObjectInstanceSelection> fireflies,
Problem prob)
{
int nFF = fireflies.Count; //number of fireflies
double rC, rG, rF; //rC -> distance for C value, rG-> distance for Gamma value, rF - distance for the feature mask
double beta0;
double beta; // beta -> attrativeness value for C and G, betaF -> attrativeness for the feature mask
//specifying the ranges for C and Gamma
double minC = Math.Pow(2, MIN_C); // minimum value for C
double maxC = Math.Pow(2, MAX_C); // maximum value for C
double minG = Math.Pow(2, MIN_G); // minimum value for G
double maxG = Math.Pow(2, MAX_G); // maximum value for G
int subsetSize = fireflies[0].Attribute_Values.Count(); //size of Instance Mask
double[] CBackup = new double[fireflies.Count]; //back up array for C value
double[] GammaBackup = new double[fireflies.Count]; ////back up array for Gamma value
double val;
Random rnd = new Random();
Random rx = new Random();
Random ry = new Random();
duplicateValue(fireflies, CBackup, GammaBackup);
for (int i = 0; i < nFF; i++)
{
for (int j = 0; j < nFF; j++)
{
if (j == i) //avoid comparism with the same element
{
continue;
}
rF = 0.0;
rC = Math.Pow(((double)fireflies[i].cValue - (double)fireflies0[j].cValue), 2);
rG = Math.Pow(((double)fireflies[i].GValue - (double)fireflies0[j].GValue), 2);
double r = Math.Sqrt(rC + rG); //r -> total distance for both C and Gamma
if (Lightn[i] < Lighto[j])
{
beta0 = 1; //setting beta to 1
beta = beta0 * Math.Exp(-gamma * Math.Pow(r, 2)); //The attractiveness parameter for C and Gamma -> beta=exp(-gamma*r)
double rand = rnd.NextDouble();
//changing firefly i position for the continuous values - i.e C and Gamma value respectively
fireflies[i].cValue = ((double)fireflies[i].cValue * (1 - beta)) + (CBackup[j] * beta) + (alpha * (rnd.NextDouble() - 0.5));
fireflies[i].GValue = ((double)fireflies[i].GValue * (1 - beta)) + (GammaBackup[j] * beta) + (alpha * (rnd.NextDouble() - 0.5));
//move the individual position of each instance mask
for (int k = 0; k < subsetSize; k++)
{
val = ((double)fireflies[i].__Attribute_Values[k] * (1 - beta)) + (GammaBackup[j] * beta) + (alpha * (rand - 0.5)); //moving position of firefly
fireflies[i].__Attribute_Values[k] = Binarize(val, rand); //convert from discrete to binary
}
findrange(fireflies[i], minC, maxC, minG, maxG); //restrict the values of C and Gamma to the specified range
}
}
//if ((double)fireflies[i].cValue != CBackup[i] || (double)fireflies[i].GValue != GammaBackup[i])
// changedIndex.Add(i); //saving the index of the firefly that has been moved for the purpose of accuracy calculation. This to reduce the number of computations
}
//calculate the new accuracy for the newly updated C and Gamma value
//ParameterSelection.Grid(prob, param, fireflies, changedIndex, avgAcc, CBackup, GammaBackup, NFOLD);
}
/// <summary>
/// generating the initial locations of n fireflies
/// </summary>
public List <ObjectInstanceSelection> init_ffa(int nFF, int subsetSize, int probSize, Problem prob)
{
Random rnd = new Random(); // Random rx = new Random(); Random ry = new Random();
List <int> rNum = Training.GetRandomNumbers(probSize, probSize); //generate N random numbers
List <ObjectInstanceSelection> attr_values = new List <ObjectInstanceSelection>();
int cnt1 = 0, cnt2 = 0, cnt3 = 0;
//create an array of size n for x and y
int[] xn = new int[subsetSize]; //instance mask
int[] pointers = new int[subsetSize]; //array contain pointer to actual individual instance represented in the instance mask
int k = 0;
for (int i = 0; i < nFF; i++)
{
xn = new int[subsetSize];
pointers = new int[subsetSize];
cnt1 = 0; cnt2 = 0; cnt3 = 0;
for (int j = 0; j < prob.Count; j++)
{
if (cnt1 < (0.7 * subsetSize) && prob.Y[j] == 1) //select 70% positive instance of the subset
{
xn[cnt3] = rnd.Next(0, 2);
pointers[cnt3] = rNum[k];
k++; cnt1++; cnt3++;
}
else if (cnt2 < (0.3 * subsetSize) && prob.Y[j] == -1)
{
xn[cnt3] = rnd.Next(0, 2);
pointers[cnt3] = rNum[k];
k++; cnt2++; cnt3++;
}
if (cnt3 >= subsetSize)
{
break;
}
}
ObjectInstanceSelection OI = new ObjectInstanceSelection(0.0, 0.0, xn, pointers);
attr_values.Add(OI);
}
return(attr_values);
}
/// <summary>
/// Main part of the Firefly Algorithm
/// </summary>
//public Problem firefly_simple(List<double> avgAcc, List<double> CValues, List<double> GValues, Problem prob)
public Problem firefly_simple(Problem prob, out double storagePercentage)
{
//int nF = 9; //number of instances
int nI = prob.X.Count(); //total number of instance in dataset
int nFF = 5; //number of fireflies. Note: NFF * subsetsize must not be greater than Size of training dataset
int subsetSize = 100; //size of each firefly Instance Mask
int MaxGeneration = 5; //number of pseudo time steps
int[] range = new int[4] {
-5, 5, -5, 5
}; //range=[xmin xmax ymin ymax]
double alpha = 0.2; //Randomness 0--1 (highly random)
double gamma = 1.0; //Absorption coefficient
int[] xn = new int[subsetSize];
double[] xo = new double[subsetSize];
double[] Lightn = new double[nFF];
double[] Lighto = new double[nFF];
double[] fitnessVal = new double[nFF];
double globalbestIntensity;
ObjectInstanceSelection globalBest = null;
//generating the initial locations of n fireflies
List <ObjectInstanceSelection> fireflies = init_ffa(nFF, subsetSize, nI, prob);
ObjectInstanceSelection[] fireflyBackup = new ObjectInstanceSelection[fireflies.Count];
ObjectInstanceSelection[] fireflyBest = new ObjectInstanceSelection[fireflies.Count];
List <int> changedIndex = new List <int>(); //changedIndex keeps track of the index of fireflies that has been moved
double newBestIntensity = new double();
int maxIndex;
bool stopSearch = false; //stopsearch is will be set to true when the a firefly with classification accuracy = 100 is found.
globalbestIntensity = double.MinValue;
//Iterations or pseudo time marching
for (int i = 0; i < MaxGeneration; i++)
{
//Evaluate objective function
fitnessVal = this.EvaluateObjectiveFunction(fireflies, prob); //evaluate objective function for each firefly
//stop searching if firefly has found the best c and G value that yields 100%
for (int t = 0; t < fitnessVal.Count(); t++)
{
//double predAccr = avgAcc[changedIndex[t]] * 100;
double predAccr = fitnessVal[t] * 100;
if (predAccr == 100) //if prediction accuracy is equal to 100, stop searching and select the firefly that gives this accuracy
{
globalBest = fireflies[changedIndex[t]];
stopSearch = true;
break;
}
}
//stop searching if firefly has found the best c and G value that yields 100%
if (stopSearch == true)
{
break;
}
//fitnessVal = this.EvaluateObjectiveFunction(fireflies, avgAcc, prob); //evaluate objective function for each firefly
newBestIntensity = fitnessVal.Max(); //get the firefly with the highest light intensity
if (newBestIntensity > globalbestIntensity)
{
globalbestIntensity = newBestIntensity;
maxIndex = Array.IndexOf(fitnessVal, newBestIntensity); //select the index for the global best
globalBest = fireflies[maxIndex]; //select the global best firefly
//bestC = (double)fireflies[maxIndex].cValue; //save the C value for the global best
//bestGamma = (double)fireflies[maxIndex].GValue; //save the Gamma for the global best
}
fireflies.CopyTo(fireflyBackup); fitnessVal.CopyTo(Lighto, 0); fitnessVal.CopyTo(Lightn, 0); //creating duplicates
//Lightn.CopyTo(Lighto, 0);
changedIndex.Clear();
ffa_move(Lightn, fireflyBackup, Lighto, alpha, gamma, fireflies, prob);
fireflies.CopyTo(fireflyBackup); //backing up the current positions of the fireflies
Lightn.CopyTo(Lighto, 0); //backing up the current intensities of the fireflies
}
//ensure that at least, 40 instances is selected for classification
int countSelected = globalBest.__Attribute_Values.Count(q => q == 1); //count the total number of selected instances
int diff, c = 0, d = 0;
int Min = 15; //minimum number of selected instances
if (countSelected < Min)
{
diff = Min - countSelected;
//if there are less than 40, add N instances, where N = the number of selected instances and 40
while (c < diff)
{
if (globalBest.__Attribute_Values[d++] == 1)
{
continue;
}
else
{
globalBest.__Attribute_Values[d++] = 1;
c++;
}
}
}
Problem subBest = buildModelMultiClass(globalBest, prob); //model for the best Instance Mast
storagePercentage = Training.StoragePercentage(subBest, prob); //calculate the percent of the original training set was retained by the reduction algorithm
return(subBest);
}
// 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;
}
/// <summary>
/// Evaluate Objective Function
/// </summary>
//public double[] EvaluateObjectiveFunction(List<ObjectInstanceSelection> fireflies, List<double> accuracy, Problem prob)
public double[] EvaluateObjectiveFunction(List <ObjectInstanceSelection> fireflies, Problem prob)
{
int NF = fireflies.Count; //NF -> number of fireflies
int tNI = fireflies.ElementAt(0).Attribute_Values.Count(); //size of each Instance Mask
double[] fitness = new double[NF];
int sum;
List <double> y = new List <double>();
List <Node[]> x = new List <Node[]>();
double C, Gamma;
for (int i = 0; i < NF; i++)
{
//building model for each instance in instance mask in each firefly object
Problem subProb = buildModel(fireflies.ElementAt(i), prob);
Parameter param = new Parameter();
if (subProb != null)
{
int countP = subProb.Y.Count(k => k == 1); //counting the total number of positive instance in the subpeoblem
int countN = subProb.Y.Count(k => k == -1); //counting the total number of negative instance in the subproblem
if (countN <= 1 || countP <= 1) //ensuring that there are at least two positive or negative instance in a subproblem
{
int m = 0;
if (countN <= 1)
{
for (int k = 0; k < prob.Count; k++) //if no negative instance, search the whole subproblem and insert two postive instance in the first and second position of subproblem
{
if (prob.Y[k] == -1)
{
subProb.X[m] = prob.X[k]; //insert negative instance in the first and second position
subProb.Y[m] = prob.Y[k]; //insert label
m++;
}
if (m == 2)
{
break;
}
}
}
else if (countP <= 1)
{
for (int k = 0; k < prob.Count; k++) //if no positive instance, search the whole subproblem and insert two postive instance in the first and second position of subproblem
{
if (prob.Y[k] == 1)
{
subProb.X[m] = prob.X[k]; //insert negative instance in the first and second position
subProb.Y[m] = prob.Y[k]; //insert label
m++;
}
if (m == 2)
{
break;
}
}
}
}
Problem subP = Training.ClusteringBoundaryInstance(subProb);
int c = subP.Count;
int count = fireflies.ElementAt(i).__Attribute_Values.Count(q => q == 1); //total number of selected instances, to be used for subsetSize
double percentageReduction = 100 * (tNI - count) / tNI; //calculating percentage reduction for each instance Mask
fitness[i] = percentageReduction;
/*
* ParameterSelection.Grid(subProb, param, "params.txt", out C, out Gamma); //select parameters for each subset
* param.C = C;
* param.Gamma = Gamma;
* Model subModel = Training.Train(subProb, param); //train each subset
* double accr = Prediction.Predict(prob, "ClassificationResults.txt", subModel, false); //use each subset to classify train dataset
* sum = 0;
* for (int j = 0; j < tNI; j++)
* sum += fireflies.ElementAt(i).Attribute_Values[j];
*
* fitness[i] = W_SVM * accr + W_Features * (double)(1 - ((double)sum / (double)tNI)); //fitness evaluation for individual firefly
* //fitness[i] = accuracy[i] + W_Features * (double)(1 - ((double)sum / (double)tNFe)); //fitness evaluation for individual firefly
*/
/*
* for (int j = 0; j < tNI; j++)
* {
* if (fireflies.ElementAt(i).__Attribute_Values[j] == 1) //if instance is selected, use for classification
* {
* int p = fireflies.ElementAt(i).__Pointers[j];
* x.Add(prob.X[p]);
* y.Add(prob.Y[p]);
* }
* else
* continue;
* }
*
* Node[][] X = new Node[x.Count][];
* double[] Y = new double[y.Count];
*
* x.CopyTo(X); //convert from list to double[] array
* y.CopyTo(Y);
*
*
* Problem subProb = new Problem(X.Count(), Y, X, X[0].GetLength(0));
*/
}
}
return(fitness);
}
//flower pollination algorithm by Yang
public Problem BinaryFlowerPollination(Problem prob, out double storagePercentage)
{
int nargin = 0, totalInstances = prob.X.Count();
int maxGeneration = 3;
int numOfFlower = 3; //population size
int subsetSize = 100; //dimension for each flower
double probabilitySwitch = 0.8; //assign probability switch
double[] flowerFitnessVal = new double[numOfFlower];
double[] newFlowerFitnessVal = new double[numOfFlower];
double globalBest = double.MinValue;
double newBest = new double();
ObjectInstanceSelection globalBestFlower = null;
int lowerBound = -2; //set lower bound - lower boundary
int upperBound = 2; //set upper bound - upper boundary
int maxIndex;
//inittalize flowers, and get global best
List <ObjectInstanceSelection> flowers = InitializeBinaryFlower(numOfFlower, subsetSize, totalInstances, prob); //initialize solution
List <ObjectInstanceSelection> newFlowers = new List <ObjectInstanceSelection>(flowers.Count); //create a clone of flowers
flowers.ForEach((item) =>
{
newFlowers.Add(new ObjectInstanceSelection(item.__Attribute_Values, item.__Attribute_Values_Continuous, item.__Pointers, item.__Fitness)); //create a clone of flowers
});
flowerFitnessVal = EvaluateObjectiveFunction(flowers, prob); //evaluate fitness value for all the flowers
newFlowerFitnessVal = EvaluateObjectiveFunction(newFlowers, prob); //evaluate fitness value for new flowers. Note: this will be the same for this function call, since pollination has not occur
FlowerFitness(flowerFitnessVal, flowers); //fitness value for each flower
FlowerFitness(newFlowerFitnessVal, newFlowers); //fitness value for new flower
globalBestFlower = EvaluateSolution(flowerFitnessVal, newFlowerFitnessVal, globalBest, flowers, newFlowers, globalBestFlower); //get the global best flower
globalBest = flowerFitnessVal.Max();
//start flower algorithm
Random r = new Random(); int x = 0;
double[] levy = new double[subsetSize];
for (int i = 0; i < maxGeneration; i++)
{
double rand = r.NextDouble();
if (rand > probabilitySwitch) //do global pollination, to produce new pollen solution
{
levy = LevyFlight(subsetSize);
for (int j = 0; j < numOfFlower; j++)
{
for (int k = 0; k < subsetSize; k++)
{
double A = levy[k] * (flowers[j].Attribute_Values[k] - globalBestFlower.Attribute_Values[k]);
double B = flowers[j].Attribute_Values[k] + A; //new pollen solution
//double A = levy[k] * (flowers[j].Attribute_Values_Continuous[k] - globalBestFlower.Attribute_Values_Continuous[k]);
//double B = flowers[j].Attribute_Values_Continuous[k] + A;
newFlowers[j].Attribute_Values[k] = ConvertToBinary(B, r.NextDouble()); //convert to binary
//newFlowers[j].__Attribute_Values[k] = TransferFunction(B, newFlowers[j].__Attribute_Values[k]); //update flower position in the binary space
}
List <int> randNum = Training.GetRandomNumbers(2, numOfFlower); //generate 2 distinct random numbers
for (int k = 0; k < subsetSize; k++)
{
double A = flowers[j].Attribute_Values[k] + (r.NextDouble() * (flowers[randNum[0]].Attribute_Values[k] - flowers[randNum[1]].Attribute_Values[k])); //randomly select two flowers from neighbourhood for pollination
//double A = flowers[j].Attribute_Values_Continuous[k] + r.NextDouble() * (flowers[randNum[0]].Attribute_Values_Continuous[k] - flowers[randNum[1]].Attribute_Values_Continuous[k]); //randomly select two flowers from neighbourhood for pollination
newFlowers[j].Attribute_Values[k] = ConvertToBinary(A, r.NextDouble()); //convert to binary
//newFlowers[j].__Attribute_Values[k] = TransferFunction(A, newFlowers[j].__Attribute_Values[k]); //update flower position in the binary space
}
}
}
else // //do local pollination, to produce new pollen solution
{
for (int j = 0; j < numOfFlower; j++)
{
List <int> randNum = Training.GetRandomNumbers(2, numOfFlower); //generate 2 distinct random numbers
for (int k = 0; k < subsetSize; k++)
{
double A = flowers[j].Attribute_Values[k] + r.NextDouble() * (flowers[randNum[0]].Attribute_Values[k] - flowers[randNum[1]].Attribute_Values[k]); //randomly select two flowers from neighbourhood for pollination
//double A = flowers[j].Attribute_Values_Continuous[k] + r.NextDouble() * (flowers[randNum[0]].Attribute_Values_Continuous[k] - flowers[randNum[1]].Attribute_Values_Continuous[k]); //randomly select two flowers from neighbourhood for pollination
newFlowers[j].Attribute_Values[k] = ConvertToBinary(A, r.NextDouble()); //convert to binary
//newFlowers[j].__Attribute_Values[k] = TransferFunction(A, newFlowers[j].__Attribute_Values[k]); //update flower position in the binary space
}
}
}
//Select best solutions from the original population and matured population for the next generation;
SelectBestSolution(flowers, newFlowers);
//evaluate new solution
newFlowerFitnessVal = EvaluateObjectiveFunction(newFlowers, prob); //evaluate fitness value for all the flowers
FlowerFitness(newFlowerFitnessVal, newFlowers); //fitness value for new flower
globalBestFlower = EvaluateSolution(flowerFitnessVal, newFlowerFitnessVal, globalBest, flowers, newFlowers, globalBestFlower); //Evaluate solution, update better solution and get global best flower
globalBest = globalBestFlower.Fitness;
//if solution has converged to a optimal user-defined point, stop search
int Max = 60; // maximum percentage reduction
if (globalBest >= Max) //if the percentage reduction has approached 60%, stop search!
{
break;
}
}
//ensure that at least, N instances are selected for classification
int min = 15; //minimum number of selected instances
globalBestFlower = AddInstances(globalBestFlower, min);
Problem subBest = fi.buildModelMultiClass(globalBestFlower, prob); //build model for the best Instance Mast
storagePercentage = Training.StoragePercentage(subBest, prob); //calculate the percent of the original training set was retained by the reduction algorithm
return(subBest);
}
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;
}
//flower pollination algorithm by Yang
public Problem FlowerPollination(Problem prob)
{
int nargin = 0, totalInstances = prob.X.Count(), maxGeneration = 500;
int numOfFlower = 10; //population size
double probabilitySwitch = 0.8; //assign probability switch
int subsetSize = 200; //dimension for each flower
double[] flowerFitnessVal = new double[numOfFlower];
double[] newFlowerFitnessVal = new double[numOfFlower];
FireflyInstanceSelection fw = new FireflyInstanceSelection();
double globalBest = double.MinValue;
double newBest = new double();
ObjectInstanceSelection globalBestFlower = null;
int lowerBound = -2; //set lower bound - lower boundary
int upperBound = 2; //set upper bound - upper boundary
int maxIndex;
//inittalize flowers, and get global best
List <ObjectInstanceSelection> flowers = InitializeFlower(numOfFlower, subsetSize, totalInstances, prob); //initialize solution
List <ObjectInstanceSelection> newFlowers = new List <ObjectInstanceSelection>(flowers.Count); //create a clone of flowers
flowers.ForEach((item) =>
{
newFlowers.Add(new ObjectInstanceSelection(item.__Attribute_Values, item.__Attribute_Values_Continuous, item.__Pointers, item.__Fitness)); //create a clone of flowers
});
flowerFitnessVal = fw.EvaluateObjectiveFunction(flowers, prob); //evaluate fitness value for all the flowers
newFlowerFitnessVal = fw.EvaluateObjectiveFunction(newFlowers, prob); //evaluate fitness value for new flowers. Note: this will be the same for this function call, since pollination has not occur
FlowerFitness(flowerFitnessVal, flowers); //fitness value for each flower
FlowerFitness(newFlowerFitnessVal, newFlowers); //fitness value for new flower
globalBestFlower = EvaluateSolution(flowerFitnessVal, newFlowerFitnessVal, globalBest, flowers, newFlowers, globalBestFlower); //get the global best flower
globalBest = flowerFitnessVal.Max();
//start flower algorithm
Random r = new Random();
double[] levy = new double[subsetSize];
for (int i = 0; i < maxGeneration; i++)
{
double rand = r.NextDouble();
if (rand > probabilitySwitch) //global pollination
{
//global pollination
for (int j = 0; j < numOfFlower; j++)
{
levy = LevyFlight(subsetSize);
for (int k = 0; k < subsetSize; k++)
{
double A = levy[k] * (flowers[j].__Attribute_Values_Continuous[k] - globalBestFlower.__Attribute_Values_Continuous[k]);
double B = flowers[j].__Attribute_Values_Continuous[k] + A;
A = SimpleBounds(B, lowerBound, upperBound); //ensure that value does not go beyond defined boundary
newFlowers[j].__Attribute_Values_Continuous[k] = A;
newFlowers[j].__Attribute_Values[k] = fw.Binarize(B, r.NextDouble()); //convert to binary
}
}
}
else //local pollination
{
for (int j = 0; j < numOfFlower; j++)
{
List <int> randNum = Training.GetRandomNumbers(2, numOfFlower); //generate 2 distinct random numbers
double epsilon = rand;
//local pollination
for (int k = 0; k < subsetSize; k++)
{
double A = flowers[j].__Attribute_Values_Continuous[k] + epsilon * (flowers[randNum[0]].__Attribute_Values_Continuous[k] - flowers[randNum[1]].__Attribute_Values_Continuous[k]); //randomly select two flowers from neighbourhood for pollination
A = SimpleBounds(A, lowerBound, upperBound); //ensure that value does not exceed defined boundary
newFlowers[j].__Attribute_Values_Continuous[k] = A; //save computation
newFlowers[j].__Attribute_Values[k] = fw.Binarize(A, r.NextDouble()); //convert to binary
}
}
}
//evaluate new solution
newFlowerFitnessVal = fw.EvaluateObjectiveFunction(newFlowers, prob); //evaluate fitness value for all the flowers
FlowerFitness(newFlowerFitnessVal, newFlowers); //fitness value for new flower
globalBestFlower = EvaluateSolution(flowerFitnessVal, newFlowerFitnessVal, globalBest, flowers, newFlowers, globalBestFlower); //Evaluate solution, update better solution and get global best flower
globalBest = flowerFitnessVal.Max();
}
//ensure that at least, 40 instances is selected for classification
int countSelected = globalBestFlower.__Attribute_Values.Count(q => q == 1); //count the total number of selected instances
int diff, c = 0, d = 0;
int Min = 40; //minimum number of selected instances
if (countSelected < Min)
{
//if there are less than N, add N instances, where N = the number of selected instances
diff = Min - countSelected;
while (c < diff)
{
if (globalBestFlower.__Attribute_Values[d++] == 1)
{
continue;
}
else
{
globalBestFlower.__Attribute_Values[d++] = 1;
c++;
}
}
}
Problem subBest = fw.buildModel(globalBestFlower, prob); //build model for the best Instance Mast
return(subBest);
}
/// <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);
}
/// <summary>
/// Evaluate Objective Function
/// </summary>
public double[] EvaluateObjectiveFunction(List <ObjectInstanceSelection> Flowers, Problem prob)
{
int NB = Flowers.Count; //NF -> number of fireflies
int tNI = Flowers.ElementAt(0).Attribute_Values.Count(); //size of each Instance Mask
double[] fitness = new double[NB];
int sum;
List <double> y = new List <double>();
List <Node[]> x = new List <Node[]>();
double C, Gamma;
for (int i = 0; i < NB; i++)
{
//building model for each instance in instance mask in each firefly object
Problem subProb = fi.buildModel(Flowers.ElementAt(i), prob);
Parameter param = new Parameter();
if (subProb != null)
{
int countP = subProb.Y.Count(k => k == 1); //counting the total number of positive instance in the subpeoblem
int countN = subProb.Y.Count(k => k == -1); //counting the total number of negative instance in the subproblem
if (countN <= 1 || countP <= 1) //ensuring that there are at least two positive or negative instance in a subproblem
{
int m = 0;
if (countN <= 1)
{
for (int k = 0; k < prob.Count; k++) //if no negative instance, search the whole subproblem and insert two postive instance in the first and second position of subproblem
{
if (prob.Y[k] == -1)
{
subProb.X[m] = prob.X[k]; //insert negative instance in the first and second position
subProb.Y[m] = prob.Y[k]; //insert label
m++;
}
if (m == 2)
{
break;
}
}
}
else if (countP <= 1)
{
for (int k = 0; k < prob.Count; k++) //if no positive instance, search the whole subproblem and insert two postive instance in the first and second position of subproblem
{
if (prob.Y[k] == 1)
{
subProb.X[m] = prob.X[k]; //insert negative instance in the first and second position
subProb.Y[m] = prob.Y[k]; //insert label
m++;
}
if (m == 2)
{
break;
}
}
}
}
Problem subP = Training.ClusteringBoundaryInstance(subProb);
int count = Flowers.ElementAt(i).Attribute_Values.Count(q => q == 1); //total number of selected instances, to be used for subsetSize
double perRedBInstances = ((double)subProb.Count / (double)subP.Count); //percentage reduction for boundary instances
double perRedFlowerInstances = (double)(tNI - count) / tNI; //percentage reduction for flower instances
//fitness[i] = (100 * perRedFlowerInstances);
fitness[i] = (100 * perRedFlowerInstances) + perRedBInstances;
//fitness[i] = 100 * ((double)count / (double)tNI);
}
}
return(fitness);
}
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>
/// generating the initial locations of n flower
/// </summary>
public List <ObjectInstanceSelection> InitializeFlower(int nFlower, int subsetSize, int probSize, Problem prob)
{
Random rnd = new Random();
List <int> rNum = Training.GetRandomNumbers(probSize, probSize); //generate N random numbers
FireflyInstanceSelection fpa = new FireflyInstanceSelection();
List <ObjectInstanceSelection> attr_values = new List <ObjectInstanceSelection>();
int cnt1 = 0, cnt2 = 0, cnt3 = 0;
//create an array of size n for x and y
int[] xn = new int[subsetSize]; //instance mask
double[] xn_Con = new double[subsetSize]; //instance mask continuous
int[] pointers = new int[subsetSize]; //array contain pointer to actual individual instance represented in the instance mask
int k = 0;
for (int i = 0; i < nFlower; i++)
{
xn = new int[subsetSize];
xn_Con = new double[subsetSize];
pointers = new int[subsetSize];
cnt1 = 0; cnt2 = 0; cnt3 = 0;
for (int j = 0; j < prob.Count; j++)
{
if (cnt1 < (0.7 * subsetSize) && prob.Y[rNum[j]] == 1) //select 70% positive instance of the subset
{
xn_Con[cnt3] = rnd.NextDouble();
xn[cnt3] = fpa.Binarize(xn_Con[cnt3], rnd.NextDouble()); //convert generated random number to binary
pointers[cnt3] = rNum[j];
k++; cnt1++; cnt3++;
}
else if (cnt2 < (0.3 * subsetSize) && prob.Y[rNum[j]] == -1)
{
xn_Con[cnt3] = rnd.NextDouble();
xn[cnt3] = fpa.Binarize(xn_Con[cnt3], rnd.NextDouble()); //convert generated random number to binary
pointers[cnt3] = rNum[j];
k++; cnt2++; cnt3++;
}
if (cnt3 >= subsetSize)
{
break;
}
}
ObjectInstanceSelection OI = new ObjectInstanceSelection(xn, xn_Con, pointers, 0.0);
attr_values.Add(OI);
}
return(attr_values);
}
///
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, float[]>(__MapToLatentFactorSpace);
}
private static void parseCommandLine(string[] args, out Parameter parameters, out Problem problem, out bool crossValidation, out int nrfold, out string modelFilename)
{
int i;
parameters = new Parameter();
// default values
crossValidation = false;
nrfold = 0;
// parse options
for (i = 0; i < args.Length; i++)
{
if (args[i][0] != '-')
{
break;
}
++i;
switch (args[i - 1][1])
{
case 's':
parameters.SvmType = (SvmType)int.Parse(args[i]);
break;
case 't':
parameters.KernelType = (KernelType)int.Parse(args[i]);
break;
case 'd':
parameters.Degree = int.Parse(args[i]);
break;
case 'g':
parameters.Gamma = double.Parse(args[i]);
break;
case 'r':
parameters.Coefficient0 = double.Parse(args[i]);
break;
case 'n':
parameters.Nu = double.Parse(args[i]);
break;
case 'm':
parameters.CacheSize = double.Parse(args[i]);
break;
case 'c':
parameters.C = double.Parse(args[i]);
break;
case 'e':
parameters.EPS = double.Parse(args[i]);
break;
case 'p':
parameters.P = double.Parse(args[i]);
break;
case 'h':
parameters.Shrinking = int.Parse(args[i]) == 1;
break;
case 'b':
parameters.Probability = int.Parse(args[i]) == 1;
break;
case 'v':
crossValidation = true;
nrfold = int.Parse(args[i]);
if (nrfold < 2)
{
throw new ArgumentException("n-fold cross validation: n must >= 2");
}
break;
case 'w':
parameters.Weights[int.Parse(args[i - 1].Substring(2))] = double.Parse(args[1]);
break;
default:
throw new ArgumentException("Unknown Parameter");
}
}
// determine filenames
if (i >= args.Length)
{
throw new ArgumentException("No input file specified");
}
problem = Problem.Read(args[i]);
if (parameters.Gamma == 0)
{
parameters.Gamma = 1.0 / problem.MaxIndex;
}
if (i < args.Length - 1)
{
modelFilename = args[i + 1];
}
else
{
int p = args[i].LastIndexOf('/') + 1;
modelFilename = args[i].Substring(p) + ".model";
}
}
/// <summary>
/// generating the initial locations of n Cuckoo
/// </summary>
public List <ObjectInstanceSelection> InitializeBinaryCuckoo(int nNests, int subsetSize, int probSize, Problem prob)
{
//Random rnd = new Random();
//List<int> rNum = Training.GetRandomNumbers(probSize, probSize); //generate N random numbers
List <ObjectInstanceSelection> attr_values = new List <ObjectInstanceSelection>();
//int cnt1 = 0, cnt2 = 0, cnt3 = 0;
//create an array of size n for x and y
Random rnd = new Random();
//List<int> rNum = Training.GetRandomNumbers(probSize, probSize); //generate N random numbers
int[] xn = new int[subsetSize]; //instance mask
double[] xn_Con = new double[subsetSize]; //instance mask continuous
//int[] pointers = new int[subsetSize]; //array contain pointer to actual individual instance represented in the instance mask
List <double> classes = fi.getClassLabels(prob.Y); //get the class labels
int nClass = classes.Count;
int div = subsetSize / nClass;
//double freq = new double(); //initialize the frequency of all the bats to zero
//double[] vel = new double[subsetSize]; //initialize the velocity of all the bats to zero
//select pointers to instances for all the particles
//int k = 0;
if (nClass > 2) //do this for multi-class problems
{
int[] pointers = Training.AssignClassPointers_MultipleClass(prob, subsetSize, probSize); //array contain pointer to actual individual instance represented in the instance mask
for (int a = 0; a < nNests; a++)
{
xn = new int[subsetSize]; //instance mask
xn_Con = new double[subsetSize]; //instance mask continuous
for (int j = 0; j < subsetSize; j++)
{
xn[j] = rnd.Next(0, 2);
}
//Training.InstanceMask_MultipleClass(prob, subsetSize, probSize, out xn); //initialize instance mask
ObjectInstanceSelection OI = new ObjectInstanceSelection(xn, xn_Con, pointers, 0.0);
attr_values.Add(OI);
}
}
else //do this for binary class problem
{
int[] pointers = Training.AssignClassPointersBinary(prob, probSize, subsetSize); //array contain pointer to actual individual instance represented in the instance mask
for (int i = 0; i < nNests; i++)
{
xn = new int[subsetSize];
xn_Con = new double[subsetSize];
//pointers = new int[subsetSize];
//cnt1 = 0; cnt2 = 0; cnt3 = 0;
for (int j = 0; j < subsetSize; j++)
{
xn[j] = rnd.Next(0, 2);
}
//Training.InstanceMask_Binary(prob, subsetSize, pointers, out xn);
ObjectInstanceSelection OI = new ObjectInstanceSelection(xn, xn_Con, pointers, 0.0);
attr_values.Add(OI);
//for (int j = 0; j < prob.Count; j++)
//{
// if (cnt1 < (0.7 * subsetSize) && prob.Y[rNum[j]] == -1) //select 70% positive instance of the subset
// {
// xn[cnt3] = rnd.Next(0, 2);
// pointers[cnt3] = rNum[j];
// k++; cnt1++; cnt3++;
// }
// else if (cnt2 < (0.3 * subsetSize) && prob.Y[rNum[j]] == 1)
// {
// xn[cnt3] = rnd.Next(0, 2);
// pointers[cnt3] = rNum[j];
// k++; cnt2++; cnt3++;
// }
// if (cnt3 >= subsetSize)
// break;
//}
}
}
return(attr_values);
}
// 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]]);
}
}
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 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);
}
// label: label name, start: begin of each class, count: #data of classes, perm: indices to the original data
// perm, length l, must be allocated before calling this subroutine
private static void svm_group_classes(Problem prob, int[] nr_class_ret, int[][] label_ret, int[][] start_ret, int[][] count_ret, int[] perm)
{
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[] data_label = new int[l];
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;
}
}
data_label[i] = j;
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;
}
}
int[] start = new int[nr_class];
start[0] = 0;
for (i = 1; i < nr_class; i++)
start[i] = start[i - 1] + count[i - 1];
for (i = 0; i < l; i++)
{
perm[start[data_label[i]]] = i;
++start[data_label[i]];
}
start[0] = 0;
for (i = 1; i < nr_class; i++)
start[i] = start[i - 1] + count[i - 1];
nr_class_ret[0] = nr_class;
label_ret[0] = label;
start_ret[0] = start;
count_ret[0] = count;
}
//
// 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;
}
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;
}
//build model for binary problems
public Problem buildModel(ObjectInstanceSelection firefly, Problem prob)
{
int tNI = firefly.Attribute_Values.Count(); //size of each Instance Mask
List <double> y = new List <double>();
List <Node[]> x = new List <Node[]>();
bool pos = false, neg = false;
//building model for each instance in instance mask in each firefly object
for (int j = 0; j < tNI; j++)
{
if (firefly.__Attribute_Values[j] == 1) //if instance is selected, use for classification
{
int p = firefly.__Pointers[j];
x.Add(prob.X[p]);
y.Add(prob.Y[p]);
if (prob.Y[p] == 1)
{
pos = true;
}
else if (prob.Y[p] == -1)
{
neg = true;
}
}
else
{
continue;
}
}
Node[][] X = new Node[x.Count][];
double[] Y = new double[y.Count];
//ensuring that the subproblem consist of both positive and negative instance
int k = 0;
int countP = y.Count(r => r == 1); //counting the total number of positive instance in the subpeoblem
int countN = y.Count(r => r == -1); //counting the total number of negative instance in the subproble
if (pos == false && neg == false) //if no instance (positive and negative) was selected, return null. Don't perform any computation
{
return(null);
}
else if (pos == false || countP <= 1) //if pos == false, then no positive instance is in the subproblem
{
for (int i = 0; i < prob.Count; i++) //if no positive instance, search the whole subproblem and insert two postive instance in the first and second position of subproblem
{
if (prob.Y[i] == 1)
{
x[k] = prob.X[i]; //insert negative instance in the first and second position
y[k] = prob.Y[i]; //insert label
k++;
}
if (k == 2)
{
break;
}
}
}
else if (neg == false || countN <= 1) //if neg == false, then no negative instance is in the subproblem
{
k = 0;
for (int i = 0; i < prob.Count; i++) //if no negative instance, search the whole subproblem and insert two negative instances in the first and second position of subproblem
{
if (prob.Y[i] == -1)
{
x[k] = prob.X[i]; //insert negative instance in the first and second position
y[k] = prob.Y[i]; //insert label
k++;
}
if (k == 2)
{
break;
}
}
}
x.CopyTo(X); //convert from list to double[] array
y.CopyTo(Y);
Problem subProb = new Problem(X.Count(), Y, X, X[0].GetLength(0));
return(subProb);
}
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);
}
/// <summary>
/// Determines the Range transform for the provided problem.
/// </summary>
/// <param name="prob">The Problem to analyze</param>
/// <param name="lowerBound">The lower bound for scaling</param>
/// <param name="upperBound">The upper bound for scaling</param>
/// <returns>The Range transform for the problem</returns>
public static RangeTransform Compute(Problem prob, double lowerBound, double upperBound)
{
double[] minVals = new double[prob.MaxIndex];
double[] maxVals = new double[prob.MaxIndex];
for (int i = 0; i < prob.MaxIndex; i++)
{
minVals[i] = double.MaxValue;
maxVals[i] = double.MinValue;
}
for (int i = 0; i < prob.Count; i++)
{
for (int j = 0; j < prob.X[i].Length; j++)
{
int index = prob.X[i][j].Index - 1;
double value = prob.X[i][j].Value;
minVals[index] = Math.Min(minVals[index], value);
maxVals[index] = Math.Max(maxVals[index], value);
}
}
for (int i = 0; i < prob.MaxIndex; i++)
{
if (minVals[i] == double.MaxValue || maxVals[i] == double.MinValue)
{
minVals[i] = 0;
maxVals[i] = 0;
}
}
return new RangeTransform(minVals, maxVals, lowerBound, upperBound);
}
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;
}
/// <summary>
/// Determines the Range transform for the provided problem. Uses the default lower and upper bounds.
/// </summary>
/// <param name="prob">The Problem to analyze</param>
/// <returns>The Range transform for the problem</returns>
public static RangeTransform Compute(Problem prob)
{
return Compute(prob, DEFAULT_LOWER_BOUND, DEFAULT_UPPER_BOUND);
}
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;
}
/// <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 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");
}
/// <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();
}
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>
/// Evaluate Objective Function
/// </summary>
//public double[] EvaluateObjectiveFunction(List<ObjectInstanceSelection> fireflies, List<double> accuracy, Problem prob)
public double[] EvaluateObjectiveFunction(List <ObjectInstanceSelection> Cuckoos, Problem prob)
{
int NF = Cuckoos.Count; //NF -> number of fireflies
int tNI = Cuckoos.ElementAt(0).Attribute_Values.Count(); //size of each Instance Mask
double[] fitness = new double[NF];
int sum;
List <double> classes = fi.getClassLabels(prob.Y); //get the class labels
int nClass = classes.Count;
List <double> y = new List <double>();
List <Node[]> x = new List <Node[]>();
double C, Gamma;
for (int i = 0; i < NF; i++)
{
//building model for each instance in instance mask in each firefly object
Problem subProb = fi.buildModel(Cuckoos.ElementAt(i), prob);
Parameter param = new Parameter();
if (subProb != null)
{
if (nClass == 2)
{
int countP = subProb.Y.Count(k => k == 1); //counting the total number of positive instance in the subpeoblem
int countN = subProb.Y.Count(k => k == -1); //counting the total number of negative instance in the subproblem
if (countN <= 1 || countP <= 1) //ensuring that there are at least two positive or negative instance in a subproblem
{
int m = 0;
if (countN <= 1)
{
for (int k = 0; k < prob.Count; k++) //if no negative instance, search the whole subproblem and insert two postive instance in the first and second position of subproblem
{
if (prob.Y[k] == -1)
{
subProb.X[m] = prob.X[k]; //insert negative instance in the first and second position
subProb.Y[m] = prob.Y[k]; //insert label
m++;
}
if (m == 2)
{
break;
}
}
}
else if (countP <= 1)
{
for (int k = 0; k < prob.Count; k++) //if no positive instance, search the whole subproblem and insert two postive instance in the first and second position of subproblem
{
if (prob.Y[k] == 1)
{
subProb.X[m] = prob.X[k]; //insert negative instance in the first and second position
subProb.Y[m] = prob.Y[k]; //insert label
m++;
}
if (m == 2)
{
break;
}
}
}
}
}
Problem subP = Training.ClusteringBoundaryInstance(subProb);
//int subProbCount = subP.Count; //number of selected boundary instances
int count = Cuckoos.ElementAt(i).__Attribute_Values.Count(q => q == 1); //total number of selected instances, to be used for subsetSize
//double percentageReduction = 100 * (tNI - count) / tNI; //calculating percentage reduction for each instance Mask
//double perRedBnstances = (double)(subProb.Count - subProbCount) / subProb.Count; //percentage reduction for boundary instances
//double perRedBInstances = (double)(subProb.Count - subP.Count) * subP.Count; //percentage reduction for boundary instances
double perRedBInstances = (double)(subProb.Count / subP.Count); //percentage reduction for boundary instances
double perRedCuckooInstances = (double)(tNI - count) / tNI; //percentage reduction for cuckoo instances
//fitness[i] = perRedCuckooInstances * 100;
fitness[i] = (100 * perRedCuckooInstances) + perRedBInstances;
//fitness[i] = 100 * ((double)count / (double)tNI);
//fitness[i] = 100 * perRedBnstances;
//fitness[i] = 100 * (perRedBnstances + perRedCuckooInstances);
//fitness[i] = (W_SeelctedBoundaryInstances * subProbCount) + (W_Instances * ((tNI - count) / tNI));
//fitness[i] = percentageReduction;
}
}
return(fitness);
}