public static NeuralNetwork LoadAndReturnNetwork <T>() where T : class
{
var problem = new ProblemDescriptorFactory().Get <T>();
var repository = new NeuralNetworkRepositoryFactory().Get();
return(repository.Load(problem));
}
public static NeuralNetwork TrainAndReturnNetwork <T>(bool saveTrainedNetwork) where T : class
{
var problem = new ProblemDescriptorFactory().Get <T>();
//set up trainer
var trainer = new NeuralNetworkTrainerFactory().Get();
trainer.OnLearningCycleComplete += (object sender, LearningCycleCompleteEventArgs e) =>
{
System.Console.Write(
"\rOutput Error: " + e.Error.ToString("00.00000000000000000", CultureInfo.InvariantCulture) + " (After " + e.Iteration.ToString("00000") + " iterations...)"
);
};
trainer.OnNetworkReconfigured += (object sender, NetworkReconfiguredEventArgs e) =>
{
var s = "";
for (var i = 0; i < e.HiddenLayers.Count; i++)
{
s += "HL" + (i + 1) + "N" + e.HiddenLayers[i] + "-";
}
s = s.Trim('-');
System.Console.WriteLine();
System.Console.WriteLine();
System.Console.WriteLine("Hidden Layers Setup: " + s);
System.Console.WriteLine();
};
//get requested problem and train network
var neuralNetwork = trainer.Train(problem);
//save if requested
if (saveTrainedNetwork)
{
new NeuralNetworkRepositoryFactory()
.Get()
.Save(neuralNetwork);
}
//report
System.Console.WriteLine();
System.Console.WriteLine();
System.Console.WriteLine(new String('=', 50));
System.Console.WriteLine("Training Completed!");
System.Console.WriteLine("Neural Network Accuracy: " + (100 - (Math.Round(neuralNetwork.NeuralNetworkError, 4) * 100)).ToString(CultureInfo.InvariantCulture) + "%");
System.Console.WriteLine("Press a key to continue...");
System.Console.ReadKey();
return(neuralNetwork);
}
