private static INeuralNetwork <double> InitializeNeuralNetworkStartingValues(INeuralNetworkBuilder <double> builder)
{
double[][] hiddenLayerValues = new double[2][];
hiddenLayerValues[0] = new double[10];
hiddenLayerValues[1] = new double[20];
for (int i = 0; i < 10; ++i)
{
hiddenLayerValues[0][i] = i;
}
for (int i = 0; i < 20; ++i)
{
hiddenLayerValues[1][i] = i;
}
INeuralNetwork <double> network = builder.InitializeNeuralNetworkWithData(null, hiddenLayerValues, null);
network.LearningRate = 1;
return(network);
}
public static void TrainModel(INeuralNetworkBuilder networkBuilder, string preparedDataPath, string modelPath,
int imageWidth, int numClasses)
{
//var device = DeviceDescriptor.GPUDevice(0);
var device = DeviceDescriptor.CPUDevice;
var imageDim = GetImageDim(imageWidth);
var buildOutput = networkBuilder.Build(device, FeatureStreamName, LabelsStreamName, ClassifierName,
numClasses, imageDim);
var imageSize = GetImageSize(imageWidth);
var preparedData = new PreparedDataInfo(preparedDataPath, FeatureStreamName, LabelsStreamName, imageSize,
numClasses, MinibatchSource.InfinitelyRepeat);
var trainer = CreateTrainer(buildOutput);
TrainModel(device, preparedData, buildOutput, trainer);
SaveModel(buildOutput, modelPath);
// validate the model
var preparedNewData = new PreparedDataInfo(preparedDataPath, FeatureStreamName, LabelsStreamName, imageSize,
numClasses, MinibatchSource.FullDataSweep);
ValidateModelWithMinibatchSource(modelPath, imageDim, numClasses, preparedNewData, FeatureStreamName, device);
}
