private static void CheckConvexity(string dsLocation = "D:\\dropbox\\Dropbox\\Master Course\\SEM-3\\ML\\CM_CUP_Datasets\\60percenttrain.txt")
{
AA1_MLP.DataManagers.CupDataManager dm = new AA1_MLP.DataManagers.CupDataManager();
DataSet trainDS = dm.LoadData(dsLocation, 10, 2, standardize: true);
var nOriginal = //ModelManager.LoadNetwork(@"C:\Users\ahmad\Documents\monks\Monks\5kitr_mo0.5_100_final_sgdnestrov_hdn100_lr0.001_reg0.001.n");
new Network(new List <Layer>()
{
new Layer(new ActivationIdentity(), true, 10),
new Layer(new ActivationTanh(), true, 100),
// new Layer(new ActivationLeakyRelu(),true,40),
new Layer(new ActivationIdentity(), false, 2),
}, false, AA1_MLP.Enums.WeightsInitialization.Xavier);
for (int i = 0; i < 100; i++)
{
Console.WriteLine(i);
List <double[]> weightValVsCost = GeneratePlot(trainDS, nOriginal, -10, 10, 0.01);
File.WriteAllLines(@"xcurve" + i + ".txt", weightValVsCost.OrderBy(s => s[0]).Select(x => string.Join(",", x)).ToArray());
}
Console.WriteLine();
}
/// <summary>
/// For outputting the final cup results
/// </summary>
private static void TrainAndPRoduceFinalResult()
{
AA1_MLP.DataManagers.CupDataManager dm = new AA1_MLP.DataManagers.CupDataManager();
DataSet trainDS = dm.LoadData(@"D:\dropbox\Dropbox\Master Course\SEM-3\ML\CM_CUP_Datasets\ML-17-PRJ lecture package-20171225\ML-CUP17-TR.csv", 10, 2, skip: 1, standardize: true);
DataSet FinalTestDS = dm.LoadData(@"D:\dropbox\Dropbox\Master Course\SEM-3\ML\CM_CUP_Datasets\ML-17-PRJ lecture package-20171225\ML-CUP17-TS.csv", 10, skip: 1, reportOsutput: false, standardize: true);
/*AdamParams passedParams = new AdamParams();
* IOptimizer trainer = new Adam();*/
GradientDescentParams passedParams = new GradientDescentParams();
Gradientdescent trainer = new Gradientdescent();
passedParams.numberOfEpochs = 5000;
passedParams.batchSize = 10;
passedParams.trainingSet = trainDS;
passedParams.learningRate = 0.001;
passedParams.regularization = Regularizations.L2;
passedParams.regularizationRate = 0.001;
passedParams.nestrov = true;
passedParams.resilient = false;
passedParams.resilientUpdateAccelerationRate = 2;
passedParams.resilientUpdateSlowDownRate = 0.5;
passedParams.momentum = 0.5;
passedParams.NumberOfHiddenUnits = 100;
passedParams.trueThreshold = null;
string path = "cupTrain" + passedParams.NumberOfHiddenUnits + "_lr" + passedParams.learningRate + "_reg" + passedParams.regularizationRate;
//building the architecture
Network n = new Network(new List <Layer>()
{
new Layer(new ActivationIdentity(), true, 10),
new Layer(new ActivationTanh(), true, passedParams.NumberOfHiddenUnits),
// new Layer(new ActivationLeakyRelu(),true,40),
new Layer(new ActivationIdentity(), false, 2),
}, false, AA1_MLP.Enums.WeightsInitialization.Xavier);
passedParams.network = n;
var watch = System.Diagnostics.Stopwatch.StartNew();
List <double[]> learningCurve = trainer.Train(passedParams);
watch.Stop();
var elapsedMs = watch.ElapsedMilliseconds;
Console.WriteLine("elapsed Time:{0} ms", elapsedMs);
File.WriteAllText(path + ".txt", string.Join("\n", learningCurve.Select(s => string.Join(",", s))));
ModelManager.SaveNetowrk(n, path + ".n");
var predictions = ModelManager.GeneratorCUP(FinalTestDS, n);
File.WriteAllText("OMG_LOC-OSM2-TS.txt", string.Join("\n", predictions.Select(s => string.Join(",", s))));
}
