A neural transducer; architectural issue

nikita6187 / i6-seq2seq

Testing out different seq2seq models in TensorFlow, and an implementation of a neural transducer.

6 stars 1 forks source link

Open vinusankars opened 5 years ago

vinusankars commented 5 years ago

The architecture in the paper suggests unidirectional RNNs. Is there any reason why you used bidirectional RNNs in neural_transducer.py ?

nikita6187 commented 5 years ago

This was an arbitrary decision, without a specific reason other than bi-directional rnns are more expressive