public static void CreateNetwork(FileOps fileOps)
{
var network = new BasicNetwork();
network.AddLayer(new BasicLayer(new ActivationLinear(),true,4));
network.AddLayer(new BasicLayer(new ActivationTANH(), true, 6));
network.AddLayer(new BasicLayer(new ActivationTANH(), true, 2));
network.Structure.FinalizeStructure();
network.Reset();
EncogDirectoryPersistence.SaveObject(fileOps.TrainedNeuralNetworkFile, network);
}
public static void Normalise(FileOps fileOps)
{
var analyst = new EncogAnalyst();
var wizard = new AnalystWizard(analyst);
wizard.Wizard(fileOps.BaseFile,true,AnalystFileFormat.DecpntComma);
var norm = new AnalystNormalizeCSV{ProduceOutputHeaders = true};
norm.Analyze(fileOps.TrainingFile, true, CSVFormat.English, analyst);
norm.Normalize(fileOps.NormalisedTrainingFile);
norm.Analyze(fileOps.EvaluationFile, true, CSVFormat.English, analyst);
norm.Normalize(fileOps.NormalisedEvaluationFile);
analyst.Save(fileOps.AnalystFile);
}
public static void Evaluate(FileOps fileOps)
{
var network = (BasicNetwork) EncogDirectoryPersistence.LoadObject(fileOps.TrainedNeuralNetworkFile);
var analyst = new EncogAnalyst();
analyst.Load(fileOps.AnalystFile);
var evaluationSet = EncogUtility.LoadCSV2Memory(fileOps.NormalisedEvaluationFile.ToString(),
network.InputCount, network.OutputCount, true, CSVFormat.English, false);
var iteration = 0;
var hitCount = 0;
foreach (var evaluation in evaluationSet)
{
iteration++;
var output = network.Compute(evaluation.Input);
var sepalL = analyst.Script.Normalize.NormalizedFields[0].DeNormalize(evaluation.Input[0]);
var sepalW = analyst.Script.Normalize.NormalizedFields[0].DeNormalize(evaluation.Input[1]);
var petalL = analyst.Script.Normalize.NormalizedFields[0].DeNormalize(evaluation.Input[2]);
var petalW = analyst.Script.Normalize.NormalizedFields[0].DeNormalize(evaluation.Input[3]);
var classCount = analyst.Script.Normalize.NormalizedFields[4].Classes.Count;
var normalisationHigh = analyst.Script.Normalize.NormalizedFields[4].NormalizedHigh;
var normalisationLow = analyst.Script.Normalize.NormalizedFields[4].NormalizedLow;
var eq = new Equilateral(classCount, normalisationHigh, normalisationLow);
var predictedClassInt = eq.Decode(output);
var predictedClass = analyst.Script.Normalize.NormalizedFields[4].Classes[predictedClassInt].Name;
var idealClassInt = eq.Decode(evaluation.Ideal);
var idealClass = analyst.Script.Normalize.NormalizedFields[4].Classes[idealClassInt].Name;
Console.WriteLine("Predicted: {0} Ideal: {1}",predictedClass,idealClass);
if (predictedClass == idealClass)
{
hitCount++;
}
}
Console.WriteLine("Total Test Count:{0}",iteration);
Console.WriteLine("Total Correct Predictions: {0}",hitCount);
Console.WriteLine("Success rate: {0}%", (((float)hitCount/(float)iteration)*100f));
Console.ReadKey();
}
public static void Segregate(FileOps fileOps)
{
var segregator = new SegregateCSV{ProduceOutputHeaders = true};
segregator.Targets.Add(new SegregateTargetPercent(fileOps.TrainingFile,75));
segregator.Targets.Add(new SegregateTargetPercent(fileOps.EvaluationFile, 25));
segregator.Analyze(fileOps.ShuffledFile,true,CSVFormat.English);
segregator.Process();
}
static void Main(string[] args)
{
var fo = new FileOps(@"C:\programming\Machine Learning\MyFirstSkynet\IrisClassifier\Data", "IrisData");
ShuffleDataFile(fo);
Segregate(fo);
Normalise(fo);
CreateNetwork(fo);
TrainNetwork(fo);
Evaluate(fo);
}
public static void TrainNetwork(FileOps fileOps)
{
var network = (BasicNetwork)EncogDirectoryPersistence.LoadObject(fileOps.TrainedNeuralNetworkFile);
var trainingSet = EncogUtility.LoadCSV2Memory(fileOps.NormalisedTrainingFile.ToString(), network.InputCount,
network.OutputCount, true, CSVFormat.English, false);
var train = new ResilientPropagation(network, trainingSet);
int epoch = 1;
do
{
train.Iteration();
Console.WriteLine("Epoch: {0} Error: {1}",epoch,train.Error);
epoch++;
} while (train.Error > 0.01);
Console.ReadKey();
EncogDirectoryPersistence.SaveObject(fileOps.TrainedNeuralNetworkFile,network);
}
public static void ShuffleDataFile(FileOps fileOps)
{
var shuffle = new ShuffleCSV{ProduceOutputHeaders = true};
shuffle.Analyze(fileOps.BaseFile,true,CSVFormat.English);
shuffle.Process(fileOps.ShuffledFile);
}
