Click or drag to resize

SVM Probability Classification
Classification with probability estimation

This is simple example of classification with probability estimation. As you can see, the main difference from probability density function (pdf) estimation algorithms (like, for example GMM), is that the probability can change too quickly if classes can be clearly separated even in the low pdf area. So SVC probability describes mostly not pdf, but the hyperplane.

ML SVMProbability 1

ML SVMProbability 2

Implementation

SVMClassificationProbability class inherits from SVMClassification and also features the following methods:

Method

Description

Performance

set random seed

Set random seed for LIBSVM classification train with probability estimation.

StaticSetRandomSeed(Int32)

classify

Classifies observation vector into one of the classes and estimate probabilities.

methodClassify(Vector, Vector)

Code Sample

The example of SVM Probability CVS Classification usage:

C#
Copy
 1public static void ProbabilitySVCTest()
 2{
 3    Console.WriteLine("********************************************************************");
 4    Console.WriteLine("ProbabilitySVCTest");
 5    Console.WriteLine("********************************************************************");
 6
 7    Matrix xz = new Matrix(200, 2);
 8    IntegerArray cl = new IntegerArray(xz.Rows);
 9
10    for (int i = 0; i < 100; ++i)
11    {
12        Double rho = i / 100.0;
13        Double theta = 100 * rho;
14        xz[i, 0] = rho * Math.Cos(theta) * 2 + 1;
15        xz[i, 1] = rho * Math.Sin(theta) + 2;
16        xz[i + 100, 0] = rho * Math.Cos(theta) - 1;
17        xz[i + 100, 1] = rho * Math.Sin(theta) * 2;
18        cl[i] = i < xz.Rows / 2 ? 1 : 2;
19    }
20
21    SVMClassificationProbability model = new SVMClassificationProbability();
22    SVMClassificationProbability.SetRandomSeed(1);
23    model.TrainCSVC(xz, cl, false);
24    model.Save("svm_b1rnd1_cpp");
25
26    Vector prob = new Vector(model.Classes.Length);
27
28    for (int ri = 0; ri < xz.Rows; ++ri)
29        model.Classify(xz.GetRow(ri), ref prob);
30    Console.WriteLine($"Model {model.Type} with kernel {model.Kernel}");
31}

See Also