StrippedModel.java

int inputSize = summonerSelectionInputModel.size();
INDArray input = Nd4j.zeros(inputSize, 621).castTo(DataType.DOUBLE);
INDArray labels = Nd4j.zeros(inputSize, 2).castTo(DataType.DOUBLE);

IntStream.range(0, inputSize).forEach(i -> {
    SuggestedWinnerV2Model summonerSelect = summonerSelectionInputModel.get(i);

    // A list of 621 0's and sparse 1's;
    List<Integer> inputVector = summonerSelect.convertInputsToVector(totalCards);
    IntStream.range(0, inputVector.size()).forEach(j ->
            input.putScalar(new int[]{i, j}, inputVector.get(j))
    );

    // A list of 2 0's or 1.
    List<Integer> labelVector = summonerSelect.convertOutputsToVector();
    IntStream.range(0, labelVector.size()).forEach(j ->
            labels.putScalar(new int[]{i, j}, labelVector.get(j))
    );
});
DataSet ds = new DataSet(input, labels)
ds.shuffle();
SplitTestAndTrain testAndTrain = ds.splitTestAndTrain(0.65);

DataSet trainingData = testAndTrain.getTrain();
DataSet testData = testAndTrain.getTest();

DataNormalization normalizer = new NormalizerStandardize();
normalizer.fit(trainingData);
normalizer.transform(trainingData);
normalizer.transform(testData);

File deckMln = new File(getWinnerV2FileName(manaCap));

MultiLayerConfiguration conf = new NeuralNetConfiguration.Builder()
.updater(new Nesterovs(LEARNING_RATE, MOMENTUM))
.seed(SEED)
.biasInit(0)
.dataType(DataType.DOUBLE)
.miniBatch(true)
.list()
.layer(new DenseLayer.Builder()
        .nIn(621)
        .nOut(623)
        .build())
.layer(new DenseLayer.Builder()
        .nIn(623)
        .nOut(300)
        .build())
.layer(new OutputLayer.Builder(new LossMCXENT())
        .nIn(300)
        .nOut(2)
        .activation(Activation.SOFTMAX)
        .build())
.build();

MultiLayerNetwork net = new MultiLayerNetwork(conf);
net.init();

for (int i = 0; i < EPOCHS; i++) {
  net.fit(trainingData); 
}

// PRINT OUT EVAL
INDArray output = net.output(testData.getFeatures()).castTo(DataType.DOUBLE);
INDArray labels = testData.getLabels().castTo(DataType.DOUBLE);

Evaluation eval = new Evaluation();
eval.eval(labels, output);
System.out.println(eval.stats());