LSTM and Gru of RNN_ Attention mechanism

LSTM: To solve the problem RNN The problem of gradient disappearance and explosion was born .

LSTM Neural structure

![ Please add a picture description ](https://img-blog.csdnimg.cn/42164c71054c413a8b90f504a96f0cc0.png

GRU: and LSTM The effect is almost the same , But the advantage is that the parameters are reduced . Neurons from 4 A into 3 individual , The input becomes 2 individual , The output becomes 1 individual .


Attention Mechanism ： Weight each output , Make useful information more useful .