public void ScreenGD(AA1_MLP.Entities.DataSet wholeSet, int k, List <double> momentums, List <double> learningRates, List <double> regularizationRates, List <int> humberOfHiddenNeurons, GradientDescentParams passedParams, int numOfEpochs)
{
//Calling the Train method of the trainer with the desired parameters
//n, ds, learningRate: .3, numberOfEpochs: 200, shuffle: false, debug: n.Debug, nestrov:false, momentum:0.9, resilient: false, resilientUpdateAccelerationRate: 0.3,
//resilientUpdateSlowDownRate: 0.1, regularization: AA1_MLP.Enums.RegularizationRates.L2, regularizationRate: 0.001, validationSet: dt, batchSize: 7
string reportsDirectory = "80SGDKFoldsReportsnonestrov";
if (Directory.Exists(reportsDirectory))
{
Directory.Delete(reportsDirectory, true);
}
Directory.CreateDirectory(reportsDirectory);
/*List<double> momentums = new List<double> { 0, 0.5 };
* List<double> learningRates = new List<double> { 0.005, 0.01 };
* List<double> regularizationRates = new List<double> { 0, 0.001 };
* List<int> humberOfHiddenNeurons = new List<int> { 100, 90, 80, 70, 60, 50, 40, 30, 20, 10 };*/
// GradientDescentParams passedParams = new GradientDescentParams();
IOptimizer trainer = new Gradientdescent();
//AdamParams passedParams = new AdamParams();
//IOptimizer trainer = new Adam();
passedParams.numberOfEpochs = numOfEpochs;
passedParams.batchSize = 10;
for (int idxnh = 0; idxnh < humberOfHiddenNeurons.Count; idxnh++)
{
for (int idxmo = 0; idxmo < momentums.Count; idxmo++)
{
for (int idxLR = 0; idxLR < learningRates.Count; idxLR++)
{
for (int idxReg = 0; idxReg < regularizationRates.Count; idxReg++)
{
int nh = humberOfHiddenNeurons[idxnh];
passedParams.learningRate = learningRates[idxLR];
passedParams.regularization = Regularizations.L2;
passedParams.regularizationRate = regularizationRates[idxReg];
passedParams.momentum = momentums[idxmo];
passedParams.NumberOfHiddenUnits = nh;
RunKFoldWithSetOfParams(wholeSet, k, passedParams, trainer, reportsDirectory);
}
}
}
}
}
private void RunKFoldWithSetOfParams(AA1_MLP.Entities.DataSet wholeSet, int k, INeuralTrainerParams passedParams, IOptimizer trainer, string reportsPath)
{
string kRunFolderName = string.Format("hdn{0}_k{1}_lr{2}_reg{3}", passedParams.NumberOfHiddenUnits, k, passedParams.learningRate, passedParams.regularizationRate);
string KRunfolderPath = Path.Combine(reportsPath, kRunFolderName);
if (passedParams is GradientDescentParams)
{
KRunfolderPath = string.Format("{0}_mo{1}", KRunfolderPath, ((GradientDescentParams)passedParams).momentum);
}
if (Directory.Exists(KRunfolderPath))
{
Directory.Delete(KRunfolderPath);
}
Directory.CreateDirectory(KRunfolderPath);
double avgMSE = 0;
double MEE = 0, MSE = 0;
int sizeOfDataFold = wholeSet.Labels.RowCount / k;
//the training set split
TrainDataset = new DataSet(
inputs: wholeSet.Inputs.SubMatrix(sizeOfDataFold,
wholeSet.Inputs.RowCount - sizeOfDataFold, 0, wholeSet.Inputs.ColumnCount),
labels: wholeSet.Labels.SubMatrix(sizeOfDataFold,
wholeSet.Labels.RowCount - sizeOfDataFold, 0, wholeSet.Labels.ColumnCount));
//the validation set
ValidationSplit = new DataSet(
inputs: wholeSet.Inputs.SubMatrix(0, sizeOfDataFold, 0, wholeSet.Inputs.ColumnCount),
labels: wholeSet.Labels.SubMatrix(0, sizeOfDataFold, 0, wholeSet.Labels.ColumnCount));
Console.WriteLine("Run number:{0}", 0);
passedParams.trainingSet = TrainDataset;
passedParams.validationSet = ValidationSplit;
var lc = RunExperiment(trainer, passedParams, out MEE, out MSE);
File.WriteAllText(Path.Combine(KRunfolderPath, "0_learningCurve.txt"), string.Join("\n", lc.Select(s => string.Join(",", s))));
avgMSE += MSE;
for (int idxdataFold = 1; idxdataFold < k - 1; idxdataFold++)
{
Console.WriteLine("Run number:{0}", idxdataFold);
//the training set split
TrainDataset = new DataSet(
inputs: wholeSet.Inputs.SubMatrix(0, idxdataFold * sizeOfDataFold, 0, wholeSet.Inputs.ColumnCount
).Stack(wholeSet.Inputs.SubMatrix(idxdataFold * sizeOfDataFold + sizeOfDataFold,
wholeSet.Inputs.RowCount - idxdataFold * sizeOfDataFold - sizeOfDataFold, 0, wholeSet.Inputs.ColumnCount)),
labels: wholeSet.Labels.SubMatrix(0, idxdataFold * sizeOfDataFold, 0, wholeSet.Labels.ColumnCount
).Stack(wholeSet.Labels.SubMatrix(idxdataFold * sizeOfDataFold + sizeOfDataFold,
wholeSet.Labels.RowCount - idxdataFold * sizeOfDataFold - sizeOfDataFold, 0, wholeSet.Labels.ColumnCount)));
//the validation set
ValidationSplit = new DataSet(
inputs: wholeSet.Inputs.SubMatrix(idxdataFold * sizeOfDataFold, sizeOfDataFold, 0, wholeSet.Inputs.ColumnCount),
labels: wholeSet.Labels.SubMatrix(idxdataFold * sizeOfDataFold, sizeOfDataFold, 0, wholeSet.Labels.ColumnCount));
passedParams.trainingSet = TrainDataset;
passedParams.validationSet = ValidationSplit;
lc = RunExperiment(trainer, passedParams, out MEE, out MSE);
File.WriteAllText(Path.Combine(KRunfolderPath, idxdataFold + "_learningCurve.txt"), string.Join("\n", lc.Select(s => string.Join(",", s))));
avgMSE += MSE;
}
//the training set split
TrainDataset = new DataSet(
inputs: wholeSet.Inputs.SubMatrix(0, (k - 1) * sizeOfDataFold, 0, wholeSet.Inputs.ColumnCount
).Stack(wholeSet.Inputs.SubMatrix((k - 1) * sizeOfDataFold + sizeOfDataFold,
wholeSet.Inputs.RowCount - (k - 1) * sizeOfDataFold - sizeOfDataFold, 0, wholeSet.Inputs.ColumnCount)),
labels: wholeSet.Labels.SubMatrix(0, (k - 1) * sizeOfDataFold, 0, wholeSet.Labels.ColumnCount
).Stack(wholeSet.Labels.SubMatrix((k - 1) * sizeOfDataFold + sizeOfDataFold,
wholeSet.Labels.RowCount - (k - 1) * sizeOfDataFold - sizeOfDataFold, 0, wholeSet.Labels.ColumnCount)));
//the validation set
ValidationSplit = new DataSet(
inputs: wholeSet.Inputs.SubMatrix((k - 1) * sizeOfDataFold, sizeOfDataFold, 0, wholeSet.Inputs.ColumnCount),
labels: wholeSet.Labels.SubMatrix((k - 1) * sizeOfDataFold, sizeOfDataFold, 0, wholeSet.Labels.ColumnCount));
Console.WriteLine("Run number:{0}", k - 1);
passedParams.trainingSet = TrainDataset;
passedParams.validationSet = ValidationSplit;
lc = RunExperiment(trainer, passedParams, out MEE, out MSE);
File.WriteAllText(Path.Combine(KRunfolderPath, (k - 1) + "_learningCurve.txt"), string.Join("\n", lc.Select(s => string.Join(",", s))));
avgMSE += MSE;
avgMSE /= k;
Console.WriteLine("Average MSE:{0}", avgMSE);
File.AppendAllLines(Path.Combine(reportsPath, "avgMSEs"), new string[] { string.Format("{0},{1}", kRunFolderName, avgMSE) });
}