private void modifyWeights(
DoubleMatrix trainingExample,
DoubleMatrix result,
DoubleMatrix output,
double learningFactor,
double momentum,
List<DoubleMatrix> previousModifications,
EvaluationContext evalCtx) {
List<DoubleMatrix> errors = countErrors(result, output);
List<Layer> layers = getLayers();
evalCtx.resetContext();
Iterator<ActivationFunction> activationFunctionIter = evalCtx.getActivationFunction();
DoubleMatrix temporalResult = trainingExample;
for (int i = 0; i < errors.size(); i++) {
DoubleMatrix error = errors.get(i);
int layerIndex = i + 1;
Layer layer = layers.get(layerIndex);
DoubleMatrix previousModification = previousModifications.get(i);
ActivationFunction activationFunc = activationFunctionIter.next();
ActivationFunctionDerivative derivative =
DerivativeFactory.getInstance().getDerivative(activationFunc);
if (layer.includeBias()) {
temporalResult = addBiasInput(temporalResult);
}
DoubleMatrix oldVal = temporalResult.dup();
temporalResult = layer.getWeights().mmul(temporalResult);
temporalResult = activationFunc.eval(temporalResult);
// dla kazdego neuronu w warstwie
for (int j = 0; j < layer.getWeights().rows; j++) {
double derVal = derivative.evaluate(temporalResult.get(j));
DoubleMatrix oldDelta = previousModification.getRow(j);
DoubleMatrix delta = oldVal.mul(derVal).mul(learningFactor).mul(error.get(j));
delta = delta.transpose();
delta = delta.add(oldDelta.mul(momentum));
previousModification.putRow(j, delta);
DoubleMatrix oldWeights = layer.getWeights().getRow(j);
DoubleMatrix newWeights = oldWeights.add(delta);
layer.getWeights().putRow(j, newWeights);
}
}
}
private void learn(
DoubleMatrix trainingExample,
DoubleMatrix result,
double learningFactor,
double momentum,
List<DoubleMatrix> previousModifications,
EvaluationContext evalCtx) {
List<Layer> layers = getLayers();
Layer outputLayer = layers.get(layers.size() - 1);
if (outputLayer.getLayerSize() != result.getRows()) {
throw new IllegalArgumentException("wrong vector size!");
}
evalCtx.resetContext();
DoubleMatrix output = evaluate(trainingExample, evalCtx);
modifyWeights(
trainingExample, result, output, learningFactor, momentum, previousModifications, evalCtx);
}
public static void main(String... args) {
// TEST
// BackpropagationNetwork network = new BackpropagationNetwork(new int[]{9, 3}, true);
//
// List<DoubleMatrix> trainingSet = Arrays.asList(
// new DoubleMatrix(new double[]{0, 0, 0, 0, 0, 0, 0, 0, 0}),
// new DoubleMatrix(new double[]{1, 1, 1, 0, 1, 0, 0, 1, 0}),
// new DoubleMatrix(new double[]{0, 1, 0, 1, 1, 1, 0, 1, 0})
// );
//
// List<DoubleMatrix> resultSet = Arrays.asList(
// new DoubleMatrix(new double[]{1, 0, 0}),
// new DoubleMatrix(new double[]{0, 1, 0}),
// new DoubleMatrix(new double[]{0, 0, 1})
// );
//
// EvaluationContext ctx = new EvaluationContext(Arrays.asList( (ActivationFunction)new
// Sigmoid()));
// network.learn(trainingSet, resultSet, 0.3, 30000, ctx);
//
// for(DoubleMatrix t : trainingSet){
// ctx.resetContext();
// network.evaluate(t, ctx).print();
// }
//
// ctx.resetContext();
// BackpropagationNetwork network = new BackpropagationNetwork(new int[]{2, 6, 1}, true);
//
// List<DoubleMatrix> trainingSet = Arrays.asList(
// new DoubleMatrix(new double[]{0, 0}),
// new DoubleMatrix(new double[]{1, 0}),
// new DoubleMatrix(new double[]{1, 1}),
// new DoubleMatrix(new double[]{0, 1})
// );
//
// List<DoubleMatrix> resultSet = Arrays.asList(
// new DoubleMatrix(new double[]{0}),
// new DoubleMatrix(new double[]{1}),
// new DoubleMatrix(new double[]{0}),
// new DoubleMatrix(new double[]{1})
// );
//
// EvaluationContext ctx = new EvaluationContext(Arrays.asList( (ActivationFunction)new
// Sigmoid(), new Sigmoid()));
// network.learn(trainingSet, resultSet, 0.4, 0.5, 20000, ctx);
//
// for(DoubleMatrix t : trainingSet){
// ctx.resetContext();
// network.evaluate(t, ctx).print();
// }
//
// ctx.resetContext();
BackpropagationNetwork network = new BackpropagationNetwork(new int[] {3, 8, 1}, true);
List<DoubleMatrix> trainingSet =
Arrays.asList(
new DoubleMatrix(new double[] {0, 0, 0}),
new DoubleMatrix(new double[] {0, 0, 1}),
new DoubleMatrix(new double[] {0, 1, 0}),
new DoubleMatrix(new double[] {0, 1, 1}),
new DoubleMatrix(new double[] {1, 0, 0}),
new DoubleMatrix(new double[] {1, 0, 1}),
new DoubleMatrix(new double[] {1, 1, 0}),
new DoubleMatrix(new double[] {1, 1, 1}));
List<DoubleMatrix> resultSet =
Arrays.asList(
new DoubleMatrix(new double[] {0}),
new DoubleMatrix(new double[] {1}),
new DoubleMatrix(new double[] {1}),
new DoubleMatrix(new double[] {0}),
new DoubleMatrix(new double[] {1}),
new DoubleMatrix(new double[] {0}),
new DoubleMatrix(new double[] {0}),
new DoubleMatrix(new double[] {1}));
EvaluationContext ctx =
new EvaluationContext(Arrays.asList((ActivationFunction) new Sigmoid(), new Sigmoid()));
network.learn(trainingSet, resultSet, 0.2, 0.1, 50000, ctx);
for (DoubleMatrix t : trainingSet) {
ctx.resetContext();
network.evaluate(t, ctx).print();
}
}
