package aima.core.learning.neural;
import java.util.ArrayList;
import java.util.List;
import aima.core.util.math.Matrix;
import aima.core.util.math.Vector;
/**
* @author Ravi Mohan
*
*/
public class LayerSensitivity {
/*
* contains sensitivity matrices and related calculations for each layer.
* Used for backprop learning
*/
private Matrix sensitivityMatrix;
private final Layer layer;
public LayerSensitivity(Layer layer) {
Matrix weightMatrix = layer.getWeightMatrix();
this.sensitivityMatrix = new Matrix(weightMatrix.getRowDimension(),
weightMatrix.getColumnDimension());
this.layer = layer;
}
public Matrix getSensitivityMatrix() {
return sensitivityMatrix;
}
public Matrix sensitivityMatrixFromErrorMatrix(Vector errorVector) {
Matrix derivativeMatrix = createDerivativeMatrix(layer
.getLastInducedField());
Matrix calculatedSensitivityMatrix = derivativeMatrix
.times(errorVector).times(-2.0);
sensitivityMatrix = calculatedSensitivityMatrix.copy();
return calculatedSensitivityMatrix;
}
public Matrix sensitivityMatrixFromSucceedingLayer(
LayerSensitivity nextLayerSensitivity) {
Layer nextLayer = nextLayerSensitivity.getLayer();
Matrix derivativeMatrix = createDerivativeMatrix(layer
.getLastInducedField());
Matrix weightTranspose = nextLayer.getWeightMatrix().transpose();
Matrix calculatedSensitivityMatrix = derivativeMatrix.times(
weightTranspose).times(
nextLayerSensitivity.getSensitivityMatrix());
sensitivityMatrix = calculatedSensitivityMatrix.copy();
return sensitivityMatrix;
}
public Layer getLayer() {
return layer;
}
//
// PRIVATE METHODS
//
private Matrix createDerivativeMatrix(Vector lastInducedField) {
List<Double> lst = new ArrayList<Double>();
for (int i = 0; i < lastInducedField.size(); i++) {
lst.add(new Double(layer.getActivationFunction().deriv(
lastInducedField.getValue(i))));
}
return Matrix.createDiagonalMatrix(lst);
}
}