static double[][] MakeAllData(int numInput, int numHidden,
int numOutput, int numRows, int seed)
{
Random rnd = new Random(seed);
int numWeights = (numInput * numHidden) + numHidden +
(numHidden * numOutput) + numOutput;
double[] weights = new double[numWeights]; // actually weights & biases
for (int i = 0; i < numWeights; ++i)
{
weights[i] = 20.0 * rnd.NextDouble() - 10.0; // [-10.0 to 10.0]
}
Console.WriteLine("Generating weights and biases:");
ShowVector(weights, 2, 10, true);
double[][] result = new double[numRows][]; // allocate return-result
for (int i = 0; i < numRows; ++i)
{
result[i] = new double[numInput + numOutput]; // 1-of-N in last column
}
NeuralNetwork gnn =
new NeuralNetwork(numInput, numHidden, numOutput); // generating NN
gnn.SetWeights(weights);
for (int r = 0; r < numRows; ++r) // for each row
{
// generate random inputs
double[] inputs = new double[numInput];
for (int i = 0; i < numInput; ++i)
{
inputs[i] = 20.0 * rnd.NextDouble() - 10.0; // [-10.0 to -10.0]
}
// compute outputs
double[] outputs = gnn.ComputeOutputs(inputs);
// translate outputs to 1-of-N
double[] oneOfN = new double[numOutput]; // all 0.0
int maxIndex = 0;
double maxValue = outputs[0];
for (int i = 0; i < numOutput; ++i)
{
if (outputs[i] > maxValue)
{
maxIndex = i;
maxValue = outputs[i];
}
}
oneOfN[maxIndex] = 1.0;
// place inputs and 1-of-N output values into curr row
int c = 0; // column into result[][]
for (int i = 0; i < numInput; ++i) // inputs
{
result[r][c++] = inputs[i];
}
for (int i = 0; i < numOutput; ++i) // outputs
{
result[r][c++] = oneOfN[i];
}
} // each row
return(result);
} // MakeAllData
static void Main(string[] args)
{
Console.WriteLine("\nBegin neural network back-propagation demo");
int numInput = 86; // number features 5
int numHidden = 30; //4
int numOutput = 1; // number of classes for Y //2
int numRows = 1000;
int seed = 1; // gives nice demo
//Console.WriteLine("\nGenerating " + numRows +
// " artificial data items with " + numInput + " features");
//double[][] allData = MakeAllData(numInput, numHidden, numOutput,
// numRows, seed);
double[][] allData = GetFeatures();
Console.WriteLine("Done");
//ShowMatrix(allData, allData.Length, 2, true);
Console.WriteLine("\nCreating train (80%) and test (20%) matrices");
double[][] trainData;
double[][] testData;
SplitTrainTest(allData, 0.80, seed, out trainData, out testData);
Console.WriteLine("Done\n");
Console.WriteLine("Training data:");
ShowMatrix(trainData, 4, 2, true);
Console.WriteLine("Test data:");
ShowMatrix(testData, 4, 2, true);
Console.WriteLine("Creating a " + numInput + "-" + numHidden +
"-" + numOutput + " neural network");
NeuralNetwork nn = new NeuralNetwork(numInput, numHidden, numOutput);
int maxEpochs = 1000;
double learnRate = 0.05;
double momentum = 0.01;
Console.WriteLine("\nSetting maxEpochs = " + maxEpochs);
Console.WriteLine("Setting learnRate = " + learnRate.ToString("F2"));
Console.WriteLine("Setting momentum = " + momentum.ToString("F2"));
Console.WriteLine("\nStarting training");
double[] weights = nn.Train(trainData, maxEpochs, learnRate, momentum);
Console.WriteLine("Done");
Console.WriteLine("\nFinal neural network model weights and biases:\n");
ShowVector(weights, 2, 10, true);
//double[] y = nn.ComputeOutputs(new double[] { 1.0, 2.0, 3.0, 4.0 });
//ShowVector(y, 3, 3, true);
double trainAcc = nn.Accuracy(trainData);
Console.WriteLine("\nFinal accuracy on training data = " +
trainAcc.ToString("F4"));
double testAcc = nn.Accuracy(testData);
Console.WriteLine("Final accuracy on test data = " +
testAcc.ToString("F4"));
Console.WriteLine("\nEnd back-propagation demo\n");
//SerializeObject<NeuralNetwork>(nn, @"C:\DEV\MachineLearning\SBERBANK\NN.xml");
//NeuralNetwork nn1 = DeSerializeObject<NeuralNetwork>(@"C:\DEV\MachineLearning\SBERBANK\serial_NN_XML.xml");
//var item = nn1.GetWeights();
//for (int i = 0; i < item.Length; i++)
//{
// Console.Write(item[i] + " | ");
//}
Console.ReadLine();
} // Main