// -------------------------------------------------------------------------------------------------------
/// <summary>
/// Creates a new ANN with the specified configuration class
/// </summary>
/// <param name="config"></param>
public AdaptiveNeuralNetwork(NetDescription config)
{
this.totalFields = config.featuresSizes.Length;
this.vigilancesRaising = config.adaptiveVigilanceRaising;
this.fuzzyReadout = config.fuzzyReadout;
this.activeFields = config.activeFields;
this.featuresSizes = config.featuresSizes;
this.temperatureOp = config.temperatureOp;
this.learningOp = config.learningOp;
this.fieldsClass = config.fieldsClass;
this.learningRate = config.learningRate;
this.learnVigilances = config.learningVigilances;
this.performVigilances = config.performingVigilances;
this.gammas = config.gammas;
this.alphas = config.alphas;
this.adaptiveVigilanceRate = config.adaptiveVigilanceRate;
activity = new double[totalFields][];
for (int field = 0; field < totalFields; field++)
{
activity[field] = new double[featuresSizes[field]];
}
inputFields = new double[totalFields][];
for (int field = 0; field < totalFields; field++)
{
inputFields[field] = new double[featuresSizes[field]];
}
predictionFields = new double[totalFields][];
for (int field = 0; field < totalFields; field++)
{
predictionFields[field] = new double[featuresSizes[field]];
}
activitySum = new double[totalFields];
neuronsConfidence = new List <double>();
neurons = new List <double[][]>();
createNeuron();
}
public void example1_oneField()
{
Console.WriteLine("***************************************************************");
Console.WriteLine("******* TEST 1 - THE ONE FIELD ANN ****************************");
Console.WriteLine("***************************************************************\n");
// ---------------------------------------------------------------------
// ------------ Creating a configuration file --------------------------
// ---------------------------------------------------------------------
NetDescription description = new NetDescription();
// ---------------------------------------------------------------------
// ------------ Configuring an one field Adaptive Neural Network -------
// ---------------------------------------------------------------------
description.activeFields = new bool[] { true }; // defines which fields are active to perform the categorization
description.adaptiveVigilanceRaising = new bool[] { true }; // defines which fields will be affected by the adaptive vigilance
description.fuzzyReadout = false; // defines if the prediction will be performed by the fuzzy ARTI readout
description.featuresSizes = new int[] { 4 }; // defines the fields configuration, their size and quantity
description.alphas = new double[] { 0.1 }; // defines each field alpha
description.gammas = new double[] { 0.5 }; // defines each field gamma
description.learningRate = new double[] { 1.0 }; // defines the learning rate for each field
description.learningVigilances = new double[] { 1.0 }; // defines the vigilance used for learning for each field
description.performingVigilances = new double[] { 0.0 }; // defines the vigilance used to perform predictions for each field
description.adaptiveVigilanceRate = 0.001; // defines how much the vigilance will be added when predicting
description.fieldsClass = new int[] { FieldTypes.STATE }; // defines each field type
description.learningOp = new int[] { NeuronLearning.ART_I }; // defines which composite operation will be used by the network when learning
description.temperatureOp = new int[] { NeuronActivation.ART_I }; // defines which composite operation will be used by the network when performing
// ---------------------------------------------------------------------
// ---------- Creating the Adaptive Neural Network ---------------------
// ---------------------------------------------------------------------
AdaptiveNeuralNetwork network = new AdaptiveNeuralNetwork(description); // creates the network
network.printNetStructure();
network.printNetworkParameters();
network.setDebug(true);
// ---------------------------------------------------------------------
// --- Performing learning operations to learn the observed activity ---
// ---------------------------------------------------------------------
// Received external stimulus, what the agent is seeing
double[] externalStimulus = new double[4];
externalStimulus[0] = 0.3;
externalStimulus[1] = 0.3;
externalStimulus[2] = 0.7;
externalStimulus[3] = 0.1;
// The field that will receive the observed stimulus
int fieldToWrite = 0;
// Inserting the observation into the network's activity vectors
network.setInputField(fieldToWrite, externalStimulus);
// Variable that tells to the network if it needs to learn the observed stimulus
bool learn = true;
// Performing a learning operation
network.prediction(learn);
// Received external stimulus, what the agent is seeing
externalStimulus[0] = 0.1;
externalStimulus[1] = 0.3;
externalStimulus[2] = 0.5;
externalStimulus[3] = 0.6;
// Inserting the observation into the network's activity vectors
network.setInputField(fieldToWrite, externalStimulus);
// Variable that tells to the network if it needs to learn the observed stimulus
learn = true;
// Performing a learning operation
network.prediction(learn);
// ---------------------------------------------------------------------
// --- Performing a prediction operation to read a neuron cluster ------
// ---------------------------------------------------------------------
// Received external stimulus, what the agent is seeing
externalStimulus[0] = 0.1;
externalStimulus[1] = 0.3;
externalStimulus[2] = 0.4;
externalStimulus[3] = 0.6;
// The field that will receive the observed stimulus
fieldToWrite = 0;
// Inserting the observation into the network's activity vectors
network.setInputField(fieldToWrite, externalStimulus);
// Variable that tells to the network if it needs to learn the observed stimulus
learn = false;
// Performing a learning operation
network.prediction(learn);
// ---------------------------------------------------------------------
// --- Reading a prediction for the field 0 ----------------------------
// ---------------------------------------------------------------------
// The field in which the read operation will be performed
int fieldToRead = 0;
// The array that will receive the prediction
double[] prediction = new double[4];
// Performing the reading operation
prediction = network.readPrediction(fieldToRead);
// ---------------------------------------------------------------------
// --- Printing the observed prediction --------------------------------
// ---------------------------------------------------------------------
// Pause
Console.Write("Press ENTER to continue...");
Console.ReadLine();
Console.WriteLine();
}
public void example2_threeFields()
{
Console.WriteLine("***************************************************************");
Console.WriteLine("******* TEST 2 - THE THREE FIELDS ANN (FALCON) ****************");
Console.WriteLine("***************************************************************\n");
// ---------------------------------------------------------------------
// ------------ Creating a configuration file --------------------------
// ---------------------------------------------------------------------
NetDescription description = new NetDescription();
// ---------------------------------------------------------------------
// ------------ Configuring an one field Adaptive Neural Network -------
// ---------------------------------------------------------------------
description.activeFields = new bool[] { true, true, false }; // defines which fields are active to perform the categorization
description.adaptiveVigilanceRaising = new bool[] { true, false, false }; // defines which fields will be affected by the adaptive vigilance
description.fuzzyReadout = true; // defines if the prediction will be performed by the fuzzy ARTI readout
description.featuresSizes = new int[] { 4, 4, 2 }; // defines the fields configuration, their size and quantity
description.alphas = new double[] { 0.1, 0.1, 0.1 }; // defines each field alpha
description.gammas = new double[] { 0.5, 0.5, 0.0 }; // defines each field gamma
description.learningRate = new double[] { 1.0, 1.0, 1.0 }; // defines the learning rate for each field
description.learningVigilances = new double[] { 1.0, 1.0, 1.0 }; // defines the vigilance used for learning for each field
description.performingVigilances = new double[] { 0.0, 0.0, 0.0 }; // defines the vigilance used to perform predictions for each field
description.adaptiveVigilanceRate = 0.001; // defines how much the vigilance will be added when predicting
description.fieldsClass = new int[] { FieldTypes.STATE, FieldTypes.ACTION, FieldTypes.REWARD }; // defines each field type
description.learningOp = new int[] { NeuronLearning.ART_I, NeuronLearning.ART_I, NeuronLearning.ART_II }; // defines which composite operation will be used by the network when learning
description.temperatureOp = new int[] { NeuronActivation.ART_I, NeuronActivation.ART_I, NeuronActivation.ART_II }; // defines which composite operation will be used by the network when performing
// ---------------------------------------------------------------------
// ---------- Creating the Adaptive Neural Network ---------------------
// ---------------------------------------------------------------------
AdaptiveNeuralNetwork network = new AdaptiveNeuralNetwork(description); // creates the network
network.printNetStructure();
network.printNetworkParameters();
network.setDebug(true);
// ---------------------------------------------------------------------
// --- Performing learning operations to learn the observed activity ---
// ---------------------------------------------------------------------
// Received external stimulus, what the agent is seeing
double[] externalStimulusA = new double[2];
double[] externalStimulusB = new double[2];
double[] externalStimulusC = new double[1];
// Define half of the input stimulus, since complement coding will be used for this test
externalStimulusA[0] = 0.3;
externalStimulusA[1] = 0.2;
// Define half of the input stimulus, since complement coding will be used for this test
externalStimulusB[0] = 0.1;
externalStimulusB[1] = 0.9;
// Define half of the input stimulus, since complement coding will be used for this test
externalStimulusC[0] = 1.0;
// The field that will receive the observed stimulus
int fieldToWriteA = 0;
int fieldToWriteB = 1;
int fieldToWriteC = 2;
// Inserting the observation into the network's activity vectors
network.setInputField(fieldToWriteA, externalStimulusA, ComplementCodingType.MIRRORED);
network.setInputField(fieldToWriteB, externalStimulusB, ComplementCodingType.MIRRORED);
network.setInputField(fieldToWriteC, externalStimulusC, ComplementCodingType.MIRRORED);
// Variable that tells to the network if it needs to learn the observed stimulus
bool learn = true;
// Performing a learning operation
network.prediction(learn);
// Define half of the input stimulus, since complement coding will be used for this test
externalStimulusA[0] = 0.1;
externalStimulusA[1] = 0.6;
// Define half of the input stimulus, since complement coding will be used for this test
externalStimulusB[0] = 0.2;
externalStimulusB[1] = 0.1;
// Define half of the input stimulus, since complement coding will be used for this test
externalStimulusC[0] = 0.45;
// The field that will receive the observed stimulus
fieldToWriteA = 0;
fieldToWriteB = 1;
fieldToWriteC = 2;
// Inserting the observation into the network's activity vectors
network.setInputField(fieldToWriteA, externalStimulusA, ComplementCodingType.MIRRORED);
network.setInputField(fieldToWriteB, externalStimulusB, ComplementCodingType.MIRRORED);
network.setInputField(fieldToWriteC, externalStimulusC, ComplementCodingType.MIRRORED);
// Variable that tells to the network if it needs to learn the observed stimulus
learn = true;
// Performing a learning operation
network.prediction(learn);
// ---------------------------------------------------------------------
// --- Performing a prediction operation to read a neuron cluster ------
// ---------------------------------------------------------------------
// Define half of the input stimulus, since complement coding will be used for this test
externalStimulusA[0] = 0.3;
externalStimulusA[1] = 0.2;
// Define half of the input stimulus, since complement coding will be used for this test
externalStimulusB[0] = 1.0;
externalStimulusB[1] = 1.0;
// Define half of the input stimulus, since complement coding will be used for this test
externalStimulusC[0] = 1.0;
// The field that will receive the observed stimulus
fieldToWriteA = 0;
fieldToWriteB = 1;
fieldToWriteC = 2;
// Inserting the observation into the network's activity vectors
network.setInputField(fieldToWriteA, externalStimulusA, ComplementCodingType.MIRRORED);
network.setInputField(fieldToWriteB, externalStimulusB, ComplementCodingType.DIRECT_ACCESS);
network.setInputField(fieldToWriteC, externalStimulusC, ComplementCodingType.DIRECT_ACCESS);
// Variable that tells to the network if it needs to learn the observed stimulus
learn = false;
// Performing a learning operation
network.prediction(learn);
// ---------------------------------------------------------------------
// --- Reading a prediction for the field 1 ----------------------------
// ---------------------------------------------------------------------
// The field in which the read operation will be performed for the ACTION field 1
int fieldToRead = 1;
// The array that will receive the prediction
double[] prediction = new double[4];
// Performing the reading operation
prediction = network.readPrediction(fieldToRead);
// ---------------------------------------------------------------------
// --- Printing the observed prediction --------------------------------
// ---------------------------------------------------------------------
// Pause
Console.Write("Press ENTER to continue...");
Console.ReadLine();
Console.WriteLine();
}