Many-to-1 and 1-to-many RNNs

[Burak Acar]

Hi,

Has anyone built a many-to-1 and 1-to-many RNNs? That would be the architecture of a translator, though I am not after a translator. My question is more about sequence tagging (many-to-one) problem.

Thanks

[Deniz Yuret]

The Knet tutorial has examples. Sequence tagging is typically one to one, each token gets a tag. If this is what you mean take a look at the rnn notebook. I think maybe you mean sequence classification?  i.e. assign a single class to the whole sequence. If so please take a look at the imdb example.

--
You received this message because you are subscribed to the Google Groups "knet-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to knet-users+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/knet-users/9ddd29ff-f1c1-4d58-88d2-bd37967a00f2%40googlegroups.com.

[Burak Acar]

I have checked rnn notebook. It's a many-to-many model. And yes I meant classifying the whole seq. I'll go over imdb example. Thanks

--

-----------------
Burak Acar, PhD

Professor,
Bogazici University, EE Dept.,
34342, Bebek, İstanbul
URL : www.vavlab.ee.boun.edu.tr