Hidden layer 2

The number of input neurons is equal to the output of hidden layer 1. Then the number of outputs is an arbitrary value. Again we set a smaller value considering very few inputs and features:

 DenseLayer hidden_layer_2 = new DenseLayer.Builder()
                .weightInit(WeightInit.XAVIER)
                .activation(Activation.RELU)
                .nIn(32).nOut(16)
                .build();