// Network Parameters
		int rngSeed = 123; // random number seed for reproducibility
		final Random rng = new Random(rngSeed);
		OptimizationAlgorithm algo = OptimizationAlgorithm.STOCHASTIC_GRADIENT_DESCENT;
		int  iterations = 1; //Number of iterations per minibatch
		String hiddenAct = "tanh";
		String outAct = "tanh";
		Updater updater = Updater.ADAGRAD;
		
		// Learning Parameters
		double rate = 4.0;
		double regularize = 0.000001;
		double dropOut = 0.0;
		int numEpochs = 3000;
		int batchSize = 2000;
		int printEvery = Xarr.length/batchSize;
		
		// Dimensions
		int features = 13;
		int lay1 = 100;
		int lay2 = 6;
		int outs = 6;
		
		final DataSet allData = new DataSet(X,y);
		final List<DataSet> list = allData.asList();
		Collections.shuffle(list, rng);
		DataSetIterator iterator = new ListDataSetIterator(list, batchSize);
		
        MultiLayerConfiguration conf = new NeuralNetConfiguration.Builder()
                .seed(rngSeed)
                .optimizationAlgo(algo)
                .iterations(iterations)
                .activation(hiddenAct)
                .weightInit(WeightInit.XAVIER_FAN_IN)
                .updater(updater)
                .regularization(true).l2(regularize).dropOut(dropOut)
                .list()
                .layer(0, new DenseLayer.Builder()
                		.learningRate(10)
                		.momentum(5)
                        .nIn(features)
                        .nOut(lay1)
                        .build())
//                .layer(1, new DenseLayer.Builder()
//                		.learningRate(3.0)
//                		.momentum(5.0)
//                        .nIn(lay1)
//                        .nOut(lay2)
//                        .build())
                .layer(1, new OutputLayer.Builder(LossFunction.SQUARED_LOSS)
                        .activation(outAct)
                        .learningRate(6)
                        .momentum(6)
                        .nIn(lay1)
                        .nOut(outs)
                        .build())
                .pretrain(false).backprop(true)
                .build();