Text this: Output parallelism and incremental learning of neural networks /