public static void Run()
{
var training = EncogUtility.LoadEGB2Memory(Config.TrainingFile);
var pattern = new FeedForwardPattern
{
InputNeurons = training.InputSize,
OutputNeurons = training.IdealSize,
ActivationFunction = new ActivationTANH()
};
var prune = new PruneIncremental(training, pattern, 100, 1, 10,
new ConsoleStatusReportable());
prune.AddHiddenLayer(5, 50);
prune.AddHiddenLayer(0, 50);
Console.WriteLine("Starting prune process");
prune.Process();
EncogDirectoryPersistence.SaveObject(Config.NetworkFile, prune.BestNetwork);
EncogUtility.SaveEGB(Config.TrainingFile, prune.Training);
}
/// <summary>
/// Create a simple feedforward neural network.
/// </summary>
/// <param name="input">The number of input neurons.</param>
/// <param name="hidden1">The number of hidden layer 1 neurons.</param>
/// <param name="hidden2">The number of hidden layer 2 neurons.</param>
/// <param name="output">The number of output neurons.</param>
/// <param name="tanh">True to use hyperbolic tangent activation function, false to
/// use the sigmoid activation function.</param>
/// <returns>The neural network.</returns>
public static BasicNetwork SimpleFeedForward(int input,
int hidden1, int hidden2, int output,
bool tanh)
{
var pattern = new FeedForwardPattern {
InputNeurons = input, OutputNeurons = output
};
if (tanh)
{
pattern.ActivationFunction = new ActivationTANH();
}
else
{
pattern.ActivationFunction = new ActivationSigmoid();
}
if (hidden1 > 0)
{
pattern.AddHiddenLayer(hidden1);
}
if (hidden2 > 0)
{
pattern.AddHiddenLayer(hidden2);
}
var network = (BasicNetwork)pattern.Generate();
network.Reset();
return(network);
}
public static void Incremental(FileInfo dataDir)
{
FileInfo file = FileUtil.CombinePath(dataDir, Config.TRAINING_FILE);
if (!file.Exists)
{
Console.WriteLine(@"Can't read file: " + file);
return;
}
IMLDataSet training = EncogUtility.LoadEGB2Memory(file);
var pattern = new FeedForwardPattern
{
InputNeurons = training.InputSize,
OutputNeurons = training.IdealSize,
ActivationFunction = new ActivationTANH()
};
var prune = new PruneIncremental(training, pattern, 100, 1, 10,
new ConsoleStatusReportable());
prune.AddHiddenLayer(5, 50);
prune.AddHiddenLayer(0, 50);
prune.Process();
EncogDirectoryPersistence.SaveObject(file, prune.BestNetwork);
}
public void Prune()
{
INeuralDataSet trainingSet = new BasicNeuralDataSet(networkInput, networkIdealOutput);
FeedForwardPattern pattern = new FeedForwardPattern();
pattern.InputNeurons = INPUT_NEURONS;
pattern.OutputNeurons = OUTPUT_NEURONS;
if (ACTIVIATION_FUNCTION == 1)
pattern.ActivationFunction = new ActivationSigmoid();
else if (ACTIVIATION_FUNCTION == 2)
pattern.ActivationFunction = new ActivationTANH();
else
throw new System.Exception("Only 2 activation functions have been impletemented.");
PruneIncremental prune = new PruneIncremental(trainingSet, pattern, 200, new ConsoleStatusReportable());
prune.AddHiddenLayer(10, 40);
prune.AddHiddenLayer(0, 30);
prune.Process();
network = prune.BestNetwork;
Console.WriteLine("Prune process complete.");
}
public static void Run()
{
Log("Loading training data");
var encog = new EncogPersistedCollection("market-training.dat", FileMode.Open);
var trainingSet = (BasicMLDataSet)encog.Find("market-training");
Log("Figuring out best system");
var pattern = new FeedForwardPattern
{
InputNeurons = trainingSet.InputSize,
OutputNeurons = trainingSet.IdealSize,
ActivationFunction = new ActivationTANH()
};
var prune = new PruneIncremental(trainingSet, pattern, 100, new ConsoleStatusReportable());
prune.AddHiddenLayer(1, 50);
prune.AddHiddenLayer(0, 50);
prune.Process();
Log("Done!!!!");
}
private static IMLMethod CreateFeedforwardNetwork(int input)
{
// construct a feedforward type network
var pattern = new FeedForwardPattern();
pattern.ActivationFunction = new ActivationSigmoid();
pattern.InputNeurons = input;
pattern.AddHiddenLayer(5);
pattern.OutputNeurons = 1;
return(pattern.Generate());
}
public static BasicNetwork CreateNetwork()
{
var pattern = new FeedForwardPattern {
InputNeurons = 3
};
pattern.AddHiddenLayer(50);
pattern.OutputNeurons = 1;
pattern.ActivationFunction = new ActivationTANH();
var network = (BasicNetwork)pattern.Generate();
network.Reset();
return(network);
}
/// <summary>
/// Creates the feedforward network.
/// </summary>
/// <param name="inputsize">The inputsize.</param>
/// <param name="outputsize">The outputsize.</param>
/// <param name="hiddenlayers">The hiddenlayers.</param>
/// <param name="hidden2layers">The hidden2layers.</param>
/// <returns></returns>
public static BasicNetwork CreateFeedforwardNetwork(int inputsize, int outputsize, int hiddenlayers, int hidden2layers)
{
// construct an Elman type network
FeedForwardPattern pattern = new FeedForwardPattern();
pattern.ActivationFunction = new ActivationTANH();
pattern.InputNeurons = inputsize;
pattern.AddHiddenLayer(hiddenlayers);
pattern.AddHiddenLayer(hidden2layers);
pattern.OutputNeurons = outputsize;
var network = pattern.Generate();
return((BasicNetwork)network);
}
private BasicNetwork CreateNetwork()
{
var pattern = new FeedForwardPattern {
InputNeurons = 1
};
for (int i = 0; i < simSettings.HiddenLayers; i++)
{
pattern.AddHiddenLayer(simSettings.HiddenLayerNeurons);
}
pattern.OutputNeurons = 1;
pattern.ActivationFunction = new ActivationTANH();
var network = (BasicNetwork)pattern.Generate();
network.Reset();
return(network);
}
public static BasicNetwork CreateNetwork(int start, int end, params int[] layers)
{
var pattern = new FeedForwardPattern {
InputNeurons = start
};
foreach (int i in layers)
{
pattern.AddHiddenLayer(i);
}
pattern.OutputNeurons = end;
pattern.ActivationFunction = new ActivationTANH();
var network = (BasicNetwork)pattern.Generate();
network.Reset();
return(network);
}
/// <see cref="INetwork.BuildFeedforwardNetwork"/>
public INetwork BuildFeedforwardNetwork()
{
if (TrainedNetworkFile.Exists)
{
return(this);
}
var pattern = new FeedForwardPattern {
ActivationFunction = new ActivationSigmoid(),
InputNeurons = InputCount,
OutputNeurons = OutputCount
};
var network = (BasicNetwork)pattern.Generate();
EncogDirectoryPersistence.SaveObject(TrainedNetworkFile, network);
TrainedNetworkFile = new FileInfo(TrainedNetworkFile.FullName);
return(this);
}
private void Step4_Pruning()
{
txtStatus.Text = "Prune the network";
var trainingSet = EncogUtility.LoadCSV2Memory(Config.NormalizedTrainingFile.ToString(),
22, 1, true, CSVFormat.English, false);
var pattern = new FeedForwardPattern()
{
InputNeurons = 22,
OutputNeurons = 1,
ActivationFunction = new ActivationTANH()
};
prune = new PruneIncremental(trainingSet, pattern, 100, 1, 10, this);
prune.AddHiddenLayer(1, 10);
prune.AddHiddenLayer(0, 10);
Logs.Clear();
pruneWorker.RunWorkerAsync();
}
public EncogMaze(MazeInfo maze, int hiddenLayers = 1, int?hiddenLayerNeurons = null)
{
this.maze = maze;
filePath = Path.Combine(Environment.CurrentDirectory, maze.Name + "." + ENCOG_FILE_EXTENSION);
if (!LoadNetwork())
{
// create new network
var pattern = new FeedForwardPattern()
{
InputNeurons = INPUT_RNEURONS, OutputNeurons = OUTPUT_RNEURONS
};
for (int i = 0; i < hiddenLayers; i++)
{
pattern.AddHiddenLayer(hiddenLayerNeurons == null ? maze.Width * maze.Height : hiddenLayerNeurons.Value);
}
pattern.ActivationFunction = new ActivationTANH();
network = (BasicNetwork)pattern.Generate();
network.Reset();
}
}
