Questions tagged [bidirectional-rnn]

I typically think of teacher forcing as optional when training an RNN. We may either: use the output of time-step $t$ as the input to time-step $t+1$ use the $(t+1)$th input as the input to time-step $t+1$ When I actually sat down to write a…

It seems to me that Seq2Seq models and Bidirectional RNNs try to do the same thing. Is that true? Also, when would you recommend one setup over another?

Bidirectional RNNs are used for generating the semantic vectors of the text at the sentence level and word level. In order to train a CNN for the classification tasks, images, and labels/outputs are required in general. If the data required is…