TrainHopfield does not calculate weight matrix correctly

Good Day

I created a Hopfield neural network using org.encog.examples.neural.hopfield.Hopfield to get started. I used a BasicNeuralDataSet consisting of BiPolarNeuralData objects using the following patterns:


final boolean[][] trainingPatterns = {
{ true, true, false, false },
{ true, false, false, false }
};


After a training iteration, the weight matrix is only adjusted for the last pattern in the neural data set.

Stepping through the code I noticed that the method TrainHopfield.convertHopfieldMatrix(final Synapse target, final Matrix delta) does not add to the weight matrix.

I changed Line 86 from this...


target.getMatrix().set(row, col, delta.get(row, col));


To this...


target.getMatrix().set(
row,
col,
delta.get(row, col) + target.getMatrix().get(row, col)
);


...and the weight matrix appears to be correct. I used the Hopfield Applet as reference. Please let me know if I'm missing something or perhaps not using the training class correctly.