1. Design the output layer using RnnOutputLayer:

new RnnOutputLayer.Builder(LossFunctions.LossFunction.MCXENT)
 .activation(Activation.SOFTMAX)
 .nIn(LSTM_LAYER_SIZE).nOut(labelCount).build()

2. Use the addLayer() method to add an output layer to the network configuration:

builder.addLayer("predictMortality", new RnnOutputLayer.Builder(LossFunctions.LossFunction.MCXENT)
 .activation(Activation.SOFTMAX)
 .nIn(LSTM_LAYER_SIZE).nOut(labelCount).build(),"L1");