yos1up / dnc Goto Github PK

@yos1up, do you think you would be able to implement the sparse version oft eh DNC mentioned here: https://deepmind.com/blog/deepmind-papers-nips-part-3/?

The full paper is here: https://arxiv.org/pdf/1610.09027v1.pdf

It would be wonderful to check this new feature in your code. By the way, I'm testing in a fork your development and it works like a charm. You can see here: https://github.com/Zeta36/DNC a new task for the generalization of sequence of sums and it's great the way it works.

Regards.

In the page 7 of the article, the author write:

we have used the following variant of the deep LSTM architecture

But in your code, you are using LSTM model defined in Chain framework.
So I wonder whether your model is the exactly representation of the model mentioned in the article.
I am not arguing with you, just want to discuss this with you.

will you please add code to use GPU to train the model, i am not familiar with chainer
or anyone can port the code to tensorflow
thanks