Seq2Seq 神经网络学习笔记

本文是对教程的部分翻译，并加入了一些自己的理解和归纳（Edit by Kevin）

---

课程中一些注意点：

• ①Sequence to Sequence Learning with Neural Networks

使用LSTM；上下文向量$z^n = (h_T^n, c_T^n)$

• ②Learning Phrase Representations using RNN Encoder-Decoder for Statistical Machine Translation

使用GRU；解码时加入不变的上下文向量(context vector z)

• ③Neural Machine Translation by Jointly Learning to Align and Translate

使用双向RNN(GRU);解码时加入Attention

• ④Packed Padded Sequences, Masking and Inference
• ⑤Convolutional Sequence to Sequence Learning
• ⑥Attention is All You Need