just scratch the surface
N vs N
The length of input and output sequence must be equal
char rnn：
http://karpathy.github.io/2015/05/21/rnn-effectiveness/
N vs 1
Sequence to the Category
1 vs N
the Category to the Sequence
N vs M
Seq2Seq
encoder:
Decoder:
Attention
the length of ‘c’ limits the model performance. Attention mechanism input different c in each different time to solve this problem.
Decoder with attention:
example:
(aij is learn from the model.)
