search

Waikato / wekaDeeplearning4j

Weka package for the Deeplearning4j java library
https://deeplearning.cms.waikato.ac.nz/
GNU General Public License v3.0
185 stars 202 forks source link

Is there a way to architect a bidirectional RNN? #59

Open clebersa opened 4 years ago

clebersa commented 4 years ago

This is not a feature suggestion per so, but a question about this awesome package for Weka.

I was wondering if it is possible to define a configuration for a RNN such as it can be considered a bidirectional RNN. Is there a way?

I have been reading the documentation for DL4J and there are some layers specified there that are explicitly defined as bidirectional. On Weka DL4J there is the truncated back propagation through time backward and forward parameters. Does setting the backward parameter as 0 make my network as unidirectional? Does setting both != 0 make my network bi-directional?

Thank you very much.

braun-steven commented 4 years ago

No, this is not yet implemented.