案例:Slot Filling
RNN
LSTM
Keras support
训练方法:BPTT
The error surface is though
RNN的其他应用
RNN v.s. Structured
RNN
Structured
可以结合起来
Deep and Structured will be the future
推荐一本教科书:《Deep Learning》《深度学习》(花书)
原文:https://www.cnblogs.com/huzheyu/p/lihongyi-ml-12.html