#隐藏到输出的连接由权重矩阵V 参数化
#隐藏到隐藏的循环连接由权重矩阵W 参数化
=Bidirectional RNN=
用于手写识别和语音识别
=LSTM=
记忆单元和这些门单元的组合，大大提升了RNN处理远距离依赖问题的能力 ，解决RNN网络收敛慢的问题。
=GRU=
[x] Speech 2 End-to-End Speech Recognition in English and Mandarin, JMLR 2016.
[x] Andrej Karpathy blog, The Unreasonable Effectiveness of Recurrent Neural Networks, http://karpathy.github.io/2015/05/21/rnn-effectiveness/.