RNNs网络的几种改进

1、LightRNN

        论文题目： LightRNN: Memory and Computation-Efficient Recurrent Neural Networks

        论文链接：  https://arxiv.org/abs/1610.09893

        Github : 

        (CNTK) https://github.com/Microsoft/CNTK/tree/master/Examples/Text/LightRNN

2、SRU 

        论文题目：Training RNNs as Fast as CNNs

        论文链接：https://arxiv.org/abs/1709.02755

        Github: 

        (PyTorch) https://github.com/taolei87/sru  

        (Chainer) https://github.com/musyoku/chainer-sru

        (TensorFlow) https://github.com/desire2020/SRU-tensorflow

3、IndRNN

        论文题目： Independently Recurrent Neural Network (IndRNN): Building A Longer and Deeper RNN

        论文链接： https://arxiv.org/abs/1803.04831

        Github:

        (Theano + Lasagne) https://github.com/Sunnydreamrain/IndRNN_Theano_Lasagne

        (TensorFlow) https://github.com/batzner/indrnn

4、Nested LSTM

        论文题目： Nested LSTMs

        论文链接：  https://arxiv.org/abs/1801.10308

        Github: 

        (TensorFlow) https://github.com/hannw/nlstm

        (Keras) https://github.com/titu1994/Nested-LSTM

5、Q-RNN

        论文题目： Quasi-Recurrent Neural Networks

        论文链接： https://arxiv.org/abs/1611.01576

        Github:

        (PyTorch) https://github.com/salesforce/pytorch-qrnn

        (PyTorch) https://github.com/JayParks/quasi-rnn

        (Keras 1.2) https://github.com/DingKe/qrnn  or

        (Keras 2.0) https://github.com/DingKe/nn_playground/tree/master/qrnn

        (Chainer) https://github.com/musyoku/chainer-qrnn        

6、Advanced-Phased-LSTM

        论文题目： Phased LSTM: Accelerating Recurrent Network Training for Long or 

                          Event-based Sequences

        论文链接：https://arxiv.org/abs/1610.09513

        Github：(TensorFlow) 

        https://github.com/jeongmincha/Advanced-Phased-LSTM/tree/master/VariablePhasedLSTM

7、ACT-RNN

        论文题目：Adaptive Computation Time for Recurrent Neural Networks

        论文链接： https://arxiv.org/abs/1603.08983v6

        Github:

        (TensorFlow) https://github.com/DeNeutoy/act-tensorflow

        (TensorFlow) https://github.com/mfigurnov/sact

8、r-LSTM / p-LSTM (带辅助损失的RNNs)

        论文题目： Learning Longer-term Dependencies in RNNs with Auxiliary Losses

        论文链接： https://arxiv.org/abs/1803.00144

        Github: (暂时没找到)

9、Advanced LSTM

        论文题目： Advanced LSTM: A Study About Better Time Dependency Modeling In 

                          Emotion Recognition

        论文链接： https://arxiv.org/abs/1710.10197

        Github: (暂时未找到)

最后

以上就是土豪画笔为你收集整理的RNNs网络的几种改进的全部内容，希望文章能够帮你解决RNNs网络的几种改进所遇到的程序开发问题。

如果觉得靠谱客网站的内容还不错，欢迎将靠谱客网站推荐给程序员好友。