public void InitMLP(int nrHidden, double learnRate) { int nrInput = 2; int nrOutput = 1; _network = new MLPNetwork(nrInput, nrHidden, nrOutput, learnRate); }
public void TestXOR() { double [][] _inputPattern = { new double[2] { 0.0, 0.0 }, new double[2] { 1.0, 0.0 }, new double[2] { 0.0, 1.0 }, new double[2] { 1.0, 1.0 } }; double[][] _targetOutputPattern = { new double[1] { 0.0 }, new double[1] { 1.0 }, new double[1] { 1.0 }, new double[1] { 0.0 } }; MLPNetwork network = new MLPNetwork(2, 5, 1, 2.8); network.InitLog("XOR-"); Random rnd = new Random(); for (int i = 0; i < 50000; i++) { //network.BatchTrain(_inputPattern, _targetOutputPattern); int index = rnd.Next(0, 4); Console.WriteLine(index); network.OnlineTrain(_inputPattern[index], _targetOutputPattern[index]); } for (int i = 0; i < _inputPattern.Length; i++) { double[] outst = network.Run(_inputPattern[i]); Console.WriteLine("XOR({0}, {1}) = {2:0.000}", _inputPattern[i][0], _inputPattern[i][1], outst[0]); } }
public void TestDigit() { double[] pattern0 = { 0, 1, 1, 1, 0, 0, 1, 0, 1, 0, 0, 1, 0, 1, 0, 0, 1, 0, 1, 0, 0, 1, 1, 1, 0 }; double[] pattern1 = { 0, 0, 1, 0, 0, 0, 1, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0 }; double[] pattern2 = { 0, 1, 1, 1, 0, 0, 0, 0, 1, 0, 0, 1, 1, 1, 0, 0, 1, 0, 0, 0, 0, 1, 1, 1, 0 }; double[] pattern3 = { 0, 1, 1, 1, 0, 0, 0, 0, 1, 0, 0, 1, 1, 1, 0, 0, 0, 0, 1, 0, 0, 1, 1, 1, 0 }; double[] pattern4 = { 0, 1, 0, 1, 0, 0, 1, 0, 1, 0, 0, 1, 1, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0 }; double[] pattern5 = { 0, 1, 1, 1, 0, 0, 1, 0, 0, 0, 0, 1, 1, 1, 0, 0, 0, 0, 1, 0, 0, 1, 1, 1, 0 }; double[] pattern6 = { 0, 1, 1, 1, 0, 0, 1, 0, 0, 0, 0, 1, 1, 1, 0, 0, 1, 0, 1, 0, 0, 1, 1, 1, 0 }; double[] pattern7 = { 0, 1, 1, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0 }; double[] pattern8 = { 0, 1, 1, 1, 0, 0, 1, 0, 1, 0, 0, 1, 1, 1, 0, 0, 1, 0, 1, 0, 0, 1, 1, 1, 0 }; double[] pattern9 = { 0, 1, 1, 1, 0, 0, 1, 0, 1, 0, 0, 1, 1, 1, 0, 0, 0, 0, 1, 0, 0, 1, 1, 1, 0 }; double[][] _inputPattern = { pattern0, pattern1, pattern2, pattern3, pattern4, pattern5, pattern6, pattern7, pattern8, pattern9 }; double[][] _targetOutputPattern = { new double[] { 0.0 }, new double[] { 0.1 }, new double[] { 0.2 }, new double[] { 0.3 }, new double[] { 0.4 }, new double[] { 0.5 }, new double[] { 0.6 }, new double[] { 0.7 }, new double[] { 0.8 }, new double[] { 0.9 } }; MLPNetwork network = new MLPNetwork(25, 25, 1, 0.8); network.InitLog("Digit-"); Random rnd = new Random(); for (int i = 0; i < 500000; i++) { //network.BatchTrain(_inputPattern, _targetOutputPattern); int index = rnd.Next(0, 10); network.OnlineTrain(_inputPattern[index], _targetOutputPattern[index]); } for (int i = 0; i < _inputPattern.Length; i++) { double[] output = network.Run(_inputPattern[i]); Console.WriteLine("{0} pattern = {1:0.000} which is digit {2}", i, output[0], (int)Math.Round(output[0] * 10)); } }