Week 2: Neural Nets

GloVe (Intrinsic word vector evaluation)：\(a:b :: c:?\) $$ d = \arg \max_i \dfrac{(x_b-x_a+x_c)^T x_i}{|x_b-x_a+x_c|} $$

RNN

解决梯度消失：Understanding LSTM Networks -- colah's blog

Machine Translation

2014: First seq2seq paper published [Sutskever et al. 2014]

2016: Google Translate switches from SMT to NMT