SVM.NET with Parallel Optimization

This SVM Library is based on the source code from: http://www.matthewajohnson.org/software/svm.html

Improvement:

When finding parameters, C and Gamma, in Grid-search algorithm using ParameterSelection.PGrid instead of the original ParameterSelection.Grid will increase the calculation speed.

Example in SVMTrainingSample

    static void Main(string[] args)
    {

        // Loading training and testing data 
        Problem train;
        Problem test;
        try
        {
            train = Problem.Read("SampleFiles/TrainData.txt"); //  Your Tranining Data Path
            test = Problem.Read("SampleFiles/TestData.txt");   //  Your Test Data Path

        }
        catch
        {
            Console.WriteLine("Could not find the tranning and testing files");
            return;
        }

        // Initialize parameters of the kernal function
        Parameter parameters = new Parameter();
        double C = 0;
        double Gamma = 0;


        //This will do a grid-search optimization in sequential way to find the best parameters. 

        Console.WriteLine("Searching optimal parameters of RBF in sequential way...");

        var SequentialSW = Stopwatch.StartNew();   // Stopwatch for performance evaluation
        string SeqSearchReOutputPath = null;       // Without storing the search records  
        //string SeqentailParaOutputPath = "ParaAcc_S.txt";
        ParameterSelection.Grid(train, parameters, SeqSearchReOutputPath, out C, out Gamma);
     
        SequentialSW.Stop();

        parameters.C = C;
        parameters.Gamma = Gamma;

        //This will do a grid-search optimization in parallel way to find the best parameters
       
        Console.WriteLine("Searching optimal parameters of RBF in parallel way...");

        var ParallelSW = Stopwatch.StartNew();   // Stopwatch for performance evaluation
        string ParaSearchReOutputPath = null;    // Without storing the search records  
        //string ParaSearchReOutputPath = "ParaAcc_P.txt";
        ParameterSelection.PGrid(train, parameters, ParaSearchReOutputPath, out C, out Gamma);
        ParallelSW.Stop();

        parameters.C = C;
        parameters.Gamma = Gamma;

        //Output the performance comparsion
        string SequentialElapse = SequentialSW.Elapsed.TotalSeconds.ToString();
        string ParallelElapse = ParallelSW.Elapsed.TotalSeconds.ToString();
  
        Console.WriteLine(""); 
        Console.WriteLine("Sequential Grid-Search: " + SequentialElapse + " Seconds");
        Console.WriteLine("  Parallel Grid-Search: " + ParallelElapse +   " Seconds");
       
        Console.WriteLine(""); 
        //Train the model using the optimal parameters.
        Model model = Training.Train(train, parameters);
        
        //Perform classification on the test data, 
        //Putting the results in results.txt.

        Prediction.Predict(test, "results.txt", model, false);
        Console.WriteLine("Press any key to finish the program...");
        Console.Read();
    
    }

---