public static ThreeLayer copy(
Map map, int hidden, int examples, double learningRate, double initVar) {
ThreeLayer copy = random(map.dimension(), hidden, initVar, Activations.sigmoid());
MVN source = new MVN(map.dimension());
for (int i : series(examples)) {
Point x = source.generate(), y = map.map(x);
copy.train(x, y, learningRate);
}
return copy;
}
@Override
public ThreeLayer build(List<Double> parameters) {
ThreeLayer fnn = new ThreeLayer();
fnn.n = n;
fnn.h = h;
fnn.activation = activation;
fnn.stateIn = new ArrayRealVector(n + 1);
fnn.stateIn.setEntry(n, 1.0); // * bias node
fnn.stateHidden = new ArrayRealVector(h + 1);
fnn.stateHidden.setEntry(h, 1.0); // * bias node
fnn.stateOut = new ArrayRealVector(n);
fnn.weights0 = new Array2DRowRealMatrix(h, n + 1);
fnn.weights1 = new Array2DRowRealMatrix(n, h + 1);
fnn.weights0Delta = new Array2DRowRealMatrix(h, n + 1);
fnn.weights1Delta = new Array2DRowRealMatrix(n, h + 1);
fnn.gHidden = new ArrayRealVector(h);
fnn.eHiddenL = new ArrayRealVector(h + 1);
fnn.eHidden = new ArrayRealVector(h);
int c = 0;
for (int i : series(h)) // to
for (int j : series(n + 1)) // from
fnn.weights0.setEntry(i, j, parameters.get(c++));
for (int i : series(n)) // to
for (int j : series(h + 1)) // from
fnn.weights1.setEntry(i, j, parameters.get(c++));
return fnn;
}
