using System;

namespace svmdemo
{
	using netsvm;
	/// <summary>
	/// Zusammenfassung für Class1.
	/// </summary>
	class svmdemo
	{
		/// <summary>
		/// Der Haupteinstiegspunkt für die Anwendung.
		/// </summary>
		[STAThread]
		static void Main(string[] args)
		{
			Lib.minitialize();

			ClassificationProblem prob = new ClassificationProblem();

			int count=0;

			uint NUMTRAIN = 1000;
			Console.WriteLine("Generate {0} training examples for circle problem", NUMTRAIN);
			Random rg = new Random( );
			while (count < NUMTRAIN)
			{
				// random ppoint in [-2,2] x [-2,2]
				float x = (float)rg.NextDouble()*4.0f - 2.0f;
				float y = (float)rg.NextDouble()*4.0f - 2.0f;

				float dist = x*x+y*y-1;

				// point to near to decision line ?
				// --> retry
				if (Math.Abs(dist)<0.2f) continue;

				bool label = dist > 0.0f; // point inside circle

				SparseVector sv = new SparseVector();
				sv.put(0,x); sv.put(1,y);
				prob.push_back(sv, label);
				count += 1;
			}

			Console.WriteLine("\nRUN CSVM TRAINING");
			ClassifierCSVM csvm = new ClassifierCSVM(prob, new PolynomialKernel());
			CSVMTrainInfo res = csvm.train();
			Console.WriteLine("FINISHED:");
			Console.WriteLine("   Number of iterations: {0}\n   Number of SVs: {1}", res.numIter, res.numSV);

			int ok = 0, notok = 0;

			Console.WriteLine("\nTEST SVM");

			for (int i=0; i<100; ++i)
			{
				float x = (float)rg.NextDouble()*4.0f - 2.0f;
				float y = (float)rg.NextDouble()*4.0f - 2.0f;

				float dist = x*x+y*y-1;
				
				SparseVector sv = new SparseVector();
				sv.put(0,x); sv.put(1,y);

				float svm_label = csvm.classify(sv);

				if (svm_label * dist > 0)
					ok ++;
				else
				{
					Console.WriteLine("   FALSE: {0}", sv.to_str());
					notok ++;
				}
			}

			Console.WriteLine("Test showed: {0} right classifications {1} false classifications\n", ok, notok);
			Lib.mterminate();
			
		}
	}
}