Doubts in "Quaternion Recurrent Neural Networks"

[pradeepnutakki]

I have two doubts regarding this paper: https://openreview.net/pdf?id=ByMHvs0cFQ

1. In the paper, the neurons are mentioned. Is that 'QRNN_HIDDEN_SIZE' adjustment in the code to change neurons?
2. How to increase speed of computation. Is it good by varying 'QRNN_HIDDEN_SIZE' or any other parameter?

[TParcollet]

Hi there, sorry for the late reply. 1. Yes. 2. Anything that acts on the topology will slow/speed the training process.