Parallelizing Legendre Memory Unit Training

                     Narsimha Chilkuri 1 Chris Eliasmith 1 2

             Abstract                 make it possible for us to exploit resources such as the in-
   Recently, a new recurrent neural network (RNN)        ternet,1 which produces 20TB of text data each month. A
   named the Legendre Memory Unit (LMU) was           feat such as this, from the training perspective, would be
   proposed and shown to achieve state-of-the-art        unimaginabl ...