moved momentum and learning rate into BackProp Training scheme

Author: magicindian

src/aima/learning/neural/BackPropLearning.java

@@ -6,23 +6,26 @@ public class BackPropLearning implements NNTrainingScheme {
 	private final double learningRate;
 	private final double momentum;
 
-	private final Layer hiddenLayer;
-	private final Layer outputLayer;
-	private final LayerSensitivity hiddenSensitivity;
-	private final LayerSensitivity outputSensitivity;
-
-	public BackPropLearning(FeedForwardNeuralNetwork network,
-			double learningRate, double momentum) {
-
-		this.hiddenLayer = network.getHiddenLayer();
-		this.outputLayer = network.getOutputLayer();
-		hiddenSensitivity = new LayerSensitivity(hiddenLayer);
-		outputSensitivity = new LayerSensitivity(outputLayer);
+	private Layer hiddenLayer;
+	private Layer outputLayer;
+	private LayerSensitivity hiddenSensitivity;
+	private LayerSensitivity outputSensitivity;
+
+	public BackPropLearning(double learningRate, double momentum) {
+
 		this.learningRate = learningRate;
 		this.momentum = momentum;
 
 	}
 
+	public void setNeuralNetwork(FeedForwardNeuralNetwork ffnn) {
+
+		this.hiddenLayer = ffnn.getHiddenLayer();
+		this.outputLayer = ffnn.getOutputLayer();
+		this.hiddenSensitivity = new LayerSensitivity(hiddenLayer);
+		this.outputSensitivity = new LayerSensitivity(outputLayer);
+	}
+
 	public Vector processInput(FeedForwardNeuralNetwork network, Vector input) {
 
 		hiddenLayer.feedForward(input);

src/aima/learning/neural/FeedForwardNeuralNetwork.java

@@ -8,20 +8,13 @@ public class FeedForwardNeuralNetwork implements FunctionApproximator {
 	private final Layer hiddenLayer;
 	private final Layer outputLayer;
 
-	private final double learningRate, momentum;
-
-	private final NNTrainingScheme trainingScheme;
+	private NNTrainingScheme trainingScheme;
 
 	/*
-	 * constructor to be used for non testing code for now assume that config
-	 * contains learning rate, momentum parameter, and number of epochs. change
-	 * this later to accomodate varied learning schemes like early stopping
+	 * constructor to be used for non testing code.
 	 */
 	public FeedForwardNeuralNetwork(NNConfig config) {
 
-		learningRate = config.getParameterAsDouble("learning_rate");
-		momentum = config.getParameterAsDouble("momentum");
-
 		int numberOfInputNeurons = config
 				.getParameterAsInteger("number_of_inputs");
 		int numberOfHiddenNeurons = config
@@ -41,7 +34,6 @@ public FeedForwardNeuralNetwork(NNConfig config) {
 		outputLayer = new Layer(numberOfOutputNeurons, numberOfHiddenNeurons,
 				lowerLimitForWeights, upperLimitForWeights,
 				new PureLinearActivationFunction());
-		trainingScheme = new BackPropLearning(this, learningRate, momentum);
 
 	}
 
@@ -52,14 +44,13 @@ public FeedForwardNeuralNetwork(NNConfig config) {
 	 */
 	public FeedForwardNeuralNetwork(Matrix hiddenLayerWeights,
 			Vector hiddenLayerBias, Matrix outputLayerWeights,
-			Vector outputLayerBias, double learningRate, double momentum) {
-		this.learningRate = learningRate;
-		this.momentum = momentum;
+			Vector outputLayerBias) {
+
 		hiddenLayer = new Layer(hiddenLayerWeights, hiddenLayerBias,
 				new LogSigActivationFunction());
 		outputLayer = new Layer(outputLayerWeights, outputLayerBias,
 				new PureLinearActivationFunction());
-		trainingScheme = new BackPropLearning(this, learningRate, momentum);
+
 	}
 
 	public void processError(Vector error) {
@@ -104,4 +95,9 @@ public Layer getOutputLayer() {
 		return outputLayer;
 	}
 
+	public void setTrainingScheme(NNTrainingScheme trainingScheme) {
+		this.trainingScheme = trainingScheme;
+		trainingScheme.setNeuralNetwork(this);
+	}
+
 }

src/aima/learning/neural/Layer.java

@@ -5,7 +5,7 @@
 import aima.util.Util;
 
 public class Layer {
-	// vectors are represented by n* 1 matrices;
+	// vectors are represented by n * 1 matrices;
 	private final Matrix weightMatrix;
 
 	Vector biasVector, lastBiasUpdateVector;

src/aima/learning/neural/NNTrainingScheme.java

@@ -4,4 +4,6 @@ public interface NNTrainingScheme {
 	Vector processInput(FeedForwardNeuralNetwork network, Vector input);
 
 	void processError(FeedForwardNeuralNetwork network, Vector error);
+
+	void setNeuralNetwork(FeedForwardNeuralNetwork ffnn);
 }

src/aima/test/learningtest/neural/AllNeuralTests.java

@@ -7,9 +7,10 @@ public class AllNeuralTests {
 	public static Test suite() {
 		TestSuite suite = new TestSuite("All tests for NN Implementation");
 
+		suite.addTest(new TestSuite(BackPropagationTests.class));
 		suite.addTest(new TestSuite(LayerTests.class));
 		suite.addTest(new TestSuite(DataSetTests.class));
-		suite.addTest(new TestSuite(FeedForwardNeuralNetworkTests.class));
+
 		return suite;
 	}
 }

…eural/FeedForwardNeuralNetworkTests.java …ingtest/neural/BackPropagationTests.javasrc/aima/test/learningtest/neural/FeedForwardNeuralNetworkTests.java renamed to src/aima/test/learningtest/neural/BackPropagationTests.java src/aima/test/learningtest/neural/FeedForwardNeuralNetworkTests.java renamed to src/aima/test/learningtest/neural/BackPropagationTests.java

@@ -1,11 +1,12 @@
 package aima.test.learningtest.neural;
 
 import junit.framework.TestCase;
+import aima.learning.neural.BackPropLearning;
 import aima.learning.neural.FeedForwardNeuralNetwork;
 import aima.learning.neural.Vector;
 import aima.util.Matrix;
 
-public class FeedForwardNeuralNetworkTests extends TestCase {
+public class BackPropagationTests extends TestCase {
 
 	public void testFeedForwardAndBAckLoopWorks() {
 		// example 11.14 of Neural Network Design by Hagan, Demuth and Beale
@@ -34,8 +35,9 @@ public void testFeedForwardAndBAckLoopWorks() {
 		double momentumFactor = 0.0;
 		FeedForwardNeuralNetwork ffnn = new FeedForwardNeuralNetwork(
 				hiddenLayerWeightMatrix, hiddenLayerBiasVector,
-				outputLayerWeightMatrix, outputLayerBiasVector, learningRate,
-				momentumFactor);
+				outputLayerWeightMatrix, outputLayerBiasVector);
+		ffnn.setTrainingScheme(new BackPropLearning(learningRate,
+				momentumFactor));
 		ffnn.processInput(input);
 		ffnn.processError(error);
 
@@ -83,8 +85,10 @@ public void testFeedForwardAndBAckLoopWorksWithMomentum() {
 		double momentumFactor = 0.5;
 		FeedForwardNeuralNetwork ffnn = new FeedForwardNeuralNetwork(
 				hiddenLayerWeightMatrix, hiddenLayerBiasVector,
-				outputLayerWeightMatrix, outputLayerBiasVector, learningRate,
-				momentumFactor);
+				outputLayerWeightMatrix, outputLayerBiasVector);
+
+		ffnn.setTrainingScheme(new BackPropLearning(learningRate,
+				momentumFactor));
 		ffnn.processInput(input);
 		ffnn.processError(error);
 
@@ -104,5 +108,4 @@ public void testFeedForwardAndBAckLoopWorksWithMomentum() {
 		assertEquals(0.6061, outputLayerBias.getValue(0), 0.001);
 
 	}
-
 }