1 paper
Tao Lei, Yu Zhang, Sida I. Wang +2
Common recurrent neural architectures scale poorly due to the intrinsic difficulty in parallelizing their state computations. In this work, we propose the Simple Recurrent Unit (SR…