A recurrent neural network without chaos

Thomas Laurent, James von Brecht

We introduce an exceptionally simple gated recurrent neural network (RNN) that achieves performance comparable to well-known gated architectures, such as LSTMs and GRUs, on the wor…