It is an implementation of Multilayer Perceptron Neuron Network. The concepts are based on Simon Haykin (2009). Neural Networks and Learning Machines, 3rd edition. ISBN: 0-13-147139-2.
To define a MLP use:
MultilayerPerceptron mlp = new MultilayerPerceptron();
mlp.setInputLayer(10, new IdentityFunction());
mlp.addHiddenLayer(2, new SigmoidFunction());
HiddenLayer hiddenLayer = new HiddenLayer();
hiddenLayer.addNeuron(new HiddenNeuron(new SigmoidFunction()));
mlp.addHiddenLayer(hiddenLayer);
mlp.setOutputLayer(1, new ThresholdFunction(0.5));To traine that MLP use:
Backpropagation back = new Backpropagation(mlp);
/*
* Creating training data
*/
List<Integer[]> inputs = new ArrayList<Integer[]>();
List<Integer[]> outputs = new ArrayList<Integer[]>();
final int AND = 0, OR = 1, XOR = 2;
for (int i = 0; i < 2; i++)
for (int j = 0; j < 2; j++) {
inputs.add(new Integer[] {i, j, AND});
outputs.add(new Integer[] {i + j == 2 ? 1 : 0});
inputs.add(new Integer[] {i, j, OR});
outputs.add(new Integer[] {i + j > 0 ? 1 : 0});
inputs.add(new Integer[] {i, j, XOR});
outputs.add(new Integer[] {i + j == 1 ? 1 : 0});
}
back.setTraineInput(inputs);
back.setTraineOutput(outputs);
back.traine();