public static async Task <bool> ExportToCSV(string mlconfigPath, DeviceDescriptor device, string filePathExport)
{
try
{
var er = await MLEvaluator.EvaluateMLConfig(mlconfigPath, device, DataSetType.Testing, EvaluationType.Results);
if (er.Actual == null)
{
throw new Exception("Export has failed. No testing nor validation datatset to export.");
}
//
List <string> strLine = new List <string>();
//include label categorical values in the export
if (er.OutputClasses != null && er.OutputClasses.Count > 1)
{
var ss = "!#OutputClasses(";
for (int i = 0; i < er.OutputClasses.Count; i++)
{
ss += $"[{i}={er.OutputClasses[i]}],";
}
var outputClassesStr = ss.Substring(0, ss.Length - 1) + ")";
strLine.Add(outputClassesStr);
}
//make header
var headerStr = string.Join(";", er.Header);
strLine.Add(headerStr);
//prepare for saving
for (int i = 0; i < er.Actual.Count; i++)
{
strLine.Add($"{er.Actual[i].ToString(CultureInfo.InvariantCulture)};{er.Predicted[i].ToString(CultureInfo.InvariantCulture)}");
}
//store content to file
//
await Task.Run(() => File.WriteAllLines(filePathExport, strLine.ToArray()));
return(true);
}
catch (Exception)
{
throw;
}
}
/// <summary>
/// Evaluate model defined in the mlconfig file
/// </summary>
/// <param name="mlconfigPath"></param>
/// <param name="device"></param>
/// <returns></returns>
public static async Task <EvaluationResult> EvaluateMLConfig(string mlconfigPath, DeviceDescriptor device, DataSetType dsType, EvaluationType evType)
{
try
{
//define eval result
var er = new EvaluationResult();
er.OutputClasses = new List <string>()
{
""
};
er.Actual = new List <float>();
er.Predicted = new List <float>();
er.Header = new List <string>();
//Load ML configuration file
var dicMParameters = MLFactory.LoadMLConfiguration(mlconfigPath);
//add full path of model folder since model file doesn't contains any absolute path
dicMParameters.Add("root", MLFactory.GetMLConfigFolder(mlconfigPath));
// get model data paths
var dicPath = MLFactory.GetMLConfigComponentPaths(dicMParameters["paths"]);
//parse feature variables
var projectValues = dicMParameters["training"].Split(MLFactory.m_cntkSpearator, StringSplitOptions.RemoveEmptyEntries);
var modelName = MLFactory.GetParameterValue(projectValues, "TrainedModel");
var nnModelPath = Path.Combine(dicMParameters["root"], modelName);
//check if model exists
if (!MLFactory.IsFileExist(nnModelPath))
{
return(er);
}
//
var dataset = MLFactory.GetDataPath(dicMParameters, dsType);
if (string.IsNullOrEmpty(dataset) || string.IsNullOrEmpty(dataset) || dataset == " ")
{
if (dsType == DataSetType.Testing)
{
dataset = MLFactory.GetDataPath(dicMParameters, DataSetType.Validation);
}
if (string.IsNullOrEmpty(dataset) || string.IsNullOrEmpty(dataset) || dataset == " ")
{
return(er);
}
}
//get output classes in case the ml problem is classification
var strCls = dicMParameters.ContainsKey("metadata") ? dicMParameters["metadata"] : "";
var oc = MLFactory.GetOutputClasses(strCls);
if (oc != null)
{
er.OutputClasses = oc;
}
//MInibatch
var mbTypestr = MLFactory.GetParameterValue(projectValues, "Type");
MinibatchType mbType = (MinibatchType)Enum.Parse(typeof(MinibatchType), mbTypestr, true);
var mbSizetr = MLFactory.GetParameterValue(projectValues, "BatchSize");
var mf = MLFactory.CreateMLFactory(dicMParameters);
//perform evaluation
var evParams = new EvaluationParameters()
{
MinibatchSize = uint.Parse(mbSizetr),
MBSource = new MinibatchSourceEx(mbType, mf.StreamConfigurations.ToArray(), mf.InputVariables, mf.OutputVariables, dataset, null, MinibatchSource.FullDataSweep, false, 0),
Input = mf.InputVariables,
Ouptut = mf.OutputVariables,
};
//evaluate model
if (evType == EvaluationType.FeaturesOnly)
{
if (!dicMParameters.ContainsKey("metadata"))
{
throw new Exception("The result cannot be exported to Excel, since no metadata is stored in mlconfig file.");
}
var desc = MLFactory.ParseRawDataSet(dicMParameters["metadata"]);
er.Header = MLFactory.GenerateHeader(desc);
var fun = Function.Load(nnModelPath, device);
//
er.DataSet = await Task.Run(() => MLEvaluator.FeaturesAndLabels(fun, evParams, device));
return(er);
}
else if (evType == EvaluationType.Results)
{
//define header
er.Header.Add(evParams.Ouptut.First().Name + "_actual");
er.Header.Add(evParams.Ouptut.First().Name + "_predicted");
var fun = Function.Load(nnModelPath, device);
//
var result = await Task.Run(() => MLEvaluator.EvaluateFunction(fun, evParams, device));
er.Actual = result.actual.ToList();
er.Predicted = result.predicted.ToList();
if (er.OutputClasses.Count < 2 && evParams.Ouptut.First().Shape.Dimensions.Last() > 1)
{
var result1 = await Task.Run(() => MLEvaluator.EvaluateFunctionEx(fun, evParams, device));
er.ActualEx = result1.actual;
er.PredictedEx = result1.predicted;
}
return(er);
}
else if (evType == EvaluationType.ResultExtended)
{
//define header
er.Header.Add(evParams.Ouptut.First().Name + "_actual");
er.Header.Add(evParams.Ouptut.First().Name + "_predicted");
er.Actual = new List <float>();
er.Predicted = new List <float>();
er.ActualEx = new List <List <float> >();
er.PredictedEx = new List <List <float> >();
//
var fun = Function.Load(nnModelPath, device);
var resultEx = await Task.Run(() => MLEvaluator.EvaluateFunctionEx(fun, evParams, device));
//var resultEx = EvaluateFunctionEx(nnModelPath, dataPath, evParams, device);
for (int i = 0; i < resultEx.actual.Count(); i++)
{
var res1 = MLValue.GetResult(resultEx.actual[i]);
er.Actual.Add(res1);
var res2 = MLValue.GetResult(resultEx.predicted[i]);
er.Predicted.Add(res2);
}
er.ActualEx = resultEx.actual;
er.PredictedEx = resultEx.predicted;
return(er);
}
else
{
throw new Exception("Unknown evaluation type!");
}
}
catch (Exception)
{
throw;
}
}
protected virtual ProgressData progressTraining(TrainingParameters trParams, Trainer trainer,
Function network, MinibatchSourceEx mbs, int epoch, TrainingProgress progress, DeviceDescriptor device)
{
//calculate average training loss and evaluation
var mbAvgLoss = trainer.PreviousMinibatchLossAverage();
var mbAvgEval = trainer.PreviousMinibatchEvaluationAverage();
//get training dataset
double trainEval = mbAvgEval;
//sometimes when the data set is huge validation model against
// full training dataset could take time, so we can skip it by setting parameter 'FullTrainingSetEval'
if (trParams.FullTrainingSetEval)
{
var evParams = new EvaluationParameters()
{
MinibatchSize = trParams.BatchSize,
MBSource = new MinibatchSourceEx(mbs.Type, this.StreamConfigurations.ToArray(), this.InputVariables, this.OutputVariables, mbs.TrainingDataFile, null, MinibatchSource.FullDataSweep, false, 0),
Ouptut = OutputVariables,
Input = InputVariables,
};
var result = MLEvaluator.EvaluateFunction(trainer.Model(), evParams, device);
trainEval = MLEvaluator.CalculateMetrics(trainer.EvaluationFunction().Name, result.actual, result.predicted, device);
////if output has more than one dimension and when the output is not categorical but numeric with more than one value
////for now only custom mini-batch source is supported this kind of variable
//if(OutputVariables.First().Shape.Dimensions.Last() > 1 && evParams.MBSource.Type== MinibatchType.Custom)
//{
// var result1 = MLEvaluator.EvaluateFunctionEx(trainer.Model(), evParams, device);
// trainEval = MLEvaluator.CalculateMetrics(trainer.EvaluationFunction().Name, result1.actual, result1.predicted, device);
//}
//else
//{
// var result = MLEvaluator.EvaluateFunction(trainer.Model(), evParams, device);
// trainEval = MLEvaluator.CalculateMetrics(trainer.EvaluationFunction().Name, result.actual, result.predicted, device);
//}
}
string bestModelPath = m_bestModelPath;
double validEval = 0;
//in case validation data set is empty don't perform test-minibatch
if (!string.IsNullOrEmpty(mbs.ValidationDataFile))
{
var evParams = new EvaluationParameters()
{
MinibatchSize = trParams.BatchSize,
//StrmsConfig = StreamConfigurations.ToArray(),
MBSource = new MinibatchSourceEx(mbs.Type, this.StreamConfigurations.ToArray(), this.InputVariables, this.OutputVariables, mbs.ValidationDataFile, null, MinibatchSource.FullDataSweep, false, 0),
Ouptut = OutputVariables,
Input = InputVariables,
};
//
var result = MLEvaluator.EvaluateFunction(trainer.Model(), evParams, device);
validEval = MLEvaluator.CalculateMetrics(trainer.EvaluationFunction().Name, result.actual, result.predicted, device);
////if output has more than one dimension and when the output is not categorical but numeric with more than one value
////for now only custom mini-batch source is supported this kind of variable
//if (OutputVariables.First().Shape.Dimensions.Last() > 1 && evParams.MBSource.Type == MinibatchType.Custom)
//{
// var result1 = MLEvaluator.EvaluateFunctionEx(trainer.Model(), evParams, device);
// validEval = MLEvaluator.CalculateMetrics(trainer.EvaluationFunction().Name, result1.actual, result1.predicted, device);
//}
//else
//{
// var result = MLEvaluator.EvaluateFunction(trainer.Model(), evParams, device);
// validEval = MLEvaluator.CalculateMetrics(trainer.EvaluationFunction().Name, result.actual, result.predicted, device);
//}
}
//here we should decide if the current model worth to be saved into temp location
// depending of the Evaluation function which sometimes can be better if it is greater that previous (e.g. ClassificationAccuracy)
if (isBetterThanPrevious(trainEval, validEval, StatMetrics.IsGoalToMinimize(trainer.EvaluationFunction())) && trParams.SaveModelWhileTraining)
{
//save model
var strFilePath = $"{trParams.ModelTempLocation}\\model_at_{epoch}of{trParams.Epochs}_epochs_TimeSpan_{DateTime.Now.Ticks}";
if (!Directory.Exists(trParams.ModelTempLocation))
{
Directory.CreateDirectory(trParams.ModelTempLocation);
}
//save temp model
network.Save(strFilePath);
//set training and validation evaluation to previous state
m_PrevTrainingEval = trainEval;
m_PrevValidationEval = validEval;
bestModelPath = strFilePath;
var tpl = Tuple.Create <double, double, string>(trainEval, validEval, strFilePath);
m_ModelEvaluations.Add(tpl);
}
m_bestModelPath = bestModelPath;
//create progressData object
var prData = new ProgressData();
prData.EpochTotal = trParams.Epochs;
prData.EpochCurrent = epoch;
prData.EvaluationFunName = trainer.EvaluationFunction().Name;
prData.TrainEval = trainEval;
prData.ValidationEval = validEval;
prData.MinibatchAverageEval = mbAvgEval;
prData.MinibatchAverageLoss = mbAvgLoss;
//prData.BestModel = bestModelPath;
//the progress is only reported if satisfied the following condition
if (progress != null && (epoch % trParams.ProgressFrequency == 0 || epoch == 1 || epoch == trParams.Epochs))
{
//add info to the history
m_trainingHistory.Add(new Tuple <int, float, float, float, float>(epoch, (float)mbAvgLoss, (float)mbAvgEval,
(float)trainEval, (float)validEval));
//send progress
progress(prData);
//
//Console.WriteLine($"Epoch={epoch} of {trParams.Epochs} processed.");
}
//return progress data
return(prData);
}
public static async Task <List <string> > PrintPerformance(string mlconfigPath, DataSetType dsType, DeviceDescriptor device)
{
try
{
var er = await MLEvaluator.EvaluateMLConfig(mlconfigPath, device, dsType, EvaluationType.ResultExtended);
if (er.Actual == null)
{
throw new Exception("Export has failed. No testing nor validation datatset to export.");
}
var pa = MLEvaluator.CalculatePerformance(er, "Testing data");
//print performance result
var strB = new List <string>();
var problemType = pa.Classes.Count() == 1 ? "Regression" : (pa.Classes.Count() == 2 ? "Binary" : "Multiclass");
///////////////////////////////REGRESSION////////////////////////////////
strB.Add("************* ANNdotNET ********************");
strB.Add("**********************Model Performance Analysis*****************************");
strB.Add($"Model name={"ML Config Evaluation"}");
strB.Add($"Problem Type ={problemType}");
strB.Add($"DataSet Name = {pa.DatSetName}");
strB.Add(" ");
//
if (problemType == "Regression")
{
// strB.Add($"************Performance Parameters Value*************************************");
strB.Add(" ");
strB.Add($"Squared Error={pa.SE}");
strB.Add($"RMSE = {pa.RMSE}");
strB.Add($"Correlation Coefficient={pa.CORR}");
strB.Add($"Determination Coefficient={pa.DETC}");
}
else if (problemType == "Binary")
{
/////////////////////BINARY CLASS/////////////////////////////////////////////
// strB.Add($"************Performance Parameters Value*************************************");
strB.Add(" ");
strB.Add($"Positive Label={pa.Classes.First()} \t\t Negative Label={pa.Classes.Last()}");
strB.Add($" ");
//
strB.Add($"True Positive = {pa.TP} \t\t False Positive = {pa.FP}");
strB.Add($"True Negative = {pa.TN} \t\t False Negative = {pa.FN}");
strB.Add($" ");
//
strB.Add($"Accuracy = {pa.Acc}, \t\t Error = {pa.ER}");
strB.Add($"Precision= {pa.Precision}, \t\t Recall = {pa.Recall}");
strB.Add($"F1 Score= {pa.F1Score}, \t\t ");
strB.Add($" ");
strB.Add($"HSS={pa.HSS}; \t PSS={pa.PSS}");
strB.Add($" ");
//strB.Add($"* - Heideke Skill Score; \t **- Peirce Scill Score");
}
else if (problemType == "Multiclass")
{
/////////////////MULTICLASS//////////////////////////////////
//strB.Add($"************Performance Parameters Value*************************************");
strB.Add(" ");
strB.Add($"Overall Accuracy={pa.OAcc} \t\t\t Average Accuracy={pa.AAcc}");
//
strB.Add($"Micro avg. Precision = {pa.MicPrec} \t\t Macro avg. Precision = {pa.MacPrec}");
strB.Add($"Micro avg. Recall = {pa.MicRcall} \t\t Macro avg. Recall = {pa.MacRcall}");
//
strB.Add($"HSS ={pa.HSS}; \t\t\t PSS ={pa.PSS}");
strB.Add($" ");
//strB.Add($"* - Heideke Skill Score; \t **- Peirce Scill Score");
//* - Heideke Skill Score; \t **- Peirce Scill Score;
}
strB.Add($"************End of Performance Report*************************************");
return(strB);
}
catch (Exception)
{
throw;
}
}