private async void BtnTrainNN_Click(object sender, EventArgs e)
{
int epochSize = 0;
//These will be our hard-coded file names. We do this for now because program doesnt work with other files yet.
string strTrainingInFile = "train-images.idx3-ubyte";
string strTrainingLblFile = "train-labels.idx1-ubyte";
//Obtain Hyper Parameters from GUI
dblEta = double.Parse(txtEta.Text);
intNeuronCnt = int.Parse(txtNeuronCnt.Text);
intMiniBatchCnt = int.Parse(txtMiniBatch.Text);
dblEpochCount = double.Parse(txtEpoch.Text);
actFunc = txtActFunc.Text;
//Declare tempNeuralNetwork class(C++) and set all the hyper parameters
MnistWrapper.MnistWrapperClass tempNeuralNetwork = new MnistWrapper.MnistWrapperClass();
tempNeuralNetwork.SetActFunc(actFunc);
tempNeuralNetwork.SetEpochCount(dblEpochCount);
tempNeuralNetwork.SetNeuronCount(intNeuronCnt);
tempNeuralNetwork.SetEta(dblEta);
tempNeuralNetwork.SetBatchSize(intMiniBatchCnt);
//Obtain the images and labels from the given file names
Task <bool> gettingImagesTask = new Task <bool>(() => { return(tempNeuralNetwork.ReadImages(strTrainingInFile)); });
Task <bool> gettingLabelsTask = new Task <bool>(() => { return(tempNeuralNetwork.ReadLabels(strTrainingLblFile)); });
gettingImagesTask.Start();
gettingLabelsTask.Start();
Cursor.Current = Cursors.WaitCursor;
DisableActions();
GenerateLoadingImgTxt(ref tempNeuralNetwork);
Cursor.Current = Cursors.Default;
epochSize = tempNeuralNetwork.GetEpochSize();
totalImages = tempNeuralNetwork.GetTotalImages();
Task trainingTask = new Task(() => { tempNeuralNetwork.TrainNetwork(); });
trainingTask.Start();
//While the Network is training with read inputs and labels, display a loading screen
LoadingScreen currLoadingScreen = new LoadingScreen();
currLoadingScreen.Show(this);
while (!(tempNeuralNetwork.FinishedTraining()))
{
currLoadingScreen.AdjustLoadBar(tempNeuralNetwork.GetImagesRead(), totalImages, tempNeuralNetwork.GetEpochIterator(), epochSize, Mode.TRAINING);
}
await trainingTask;
currLoadingScreen.Close();
EnableActions();
PrimeNeuralNetwork = tempNeuralNetwork;
}
private async void BtnTestNeuralNetwork_Click(object sender, EventArgs e)
{
//This function will only be called if a neural network has been trained.
//Now we want to train a set of images
bool gotImages = false;
bool gotLabels = false;
string strTestingInFile = "t10k-images.idx3-ubyte";
string strTestinglblFile = "t10k-labels.idx1-ubyte";
PrimeNeuralNetwork.ResetMNIST();
PrimeNeuralNetwork.SetEpochCount(1);
Task <bool> gettingImagesTask = new Task <bool>(() => { return(PrimeNeuralNetwork.ReadImages(strTestingInFile)); });
Task <bool> gettingLabelsTask = new Task <bool>(() => { return(PrimeNeuralNetwork.ReadLabels(strTestinglblFile)); });
gettingImagesTask.Start();
gettingLabelsTask.Start();
Cursor.Current = Cursors.WaitCursor;
DisableActions();
//NOTE: the following function will only end when Images and Labels are done being read
GenerateLoadingImgTxt(ref PrimeNeuralNetwork);
gotImages = await gettingImagesTask;
gotLabels = await gettingLabelsTask;
Cursor.Current = Cursors.Default;
totalImages = PrimeNeuralNetwork.GetTotalImages();
//Since the testing images and labels have been read, we can now begin testing
Task taskTestingNetwork = new Task(() => { PrimeNeuralNetwork.TestNetwork(); });
taskTestingNetwork.Start();
//Generate a testing loading bar screen
LoadingScreen currLoadingScreen = new LoadingScreen();
currLoadingScreen.Show(this);
while (!PrimeNeuralNetwork.FinishedTesting())
{
currLoadingScreen.AdjustLoadBar(PrimeNeuralNetwork.GetImagesRead(), totalImages, 0, totalImages, Mode.TESTING);
}
await taskTestingNetwork;
currLoadingScreen.Close();
DisplayResults();
EnableActions();
BtnNext.Enabled = true;
}
