nice project

[hardmaru]

Hi Jiamings,

Nice to see an implementation of fast weights! I've also been working on one (at the bottom of https://github.com/hardmaru/supercell/blob/master/supercell.py), but unfortunately haven't done extensive testing yet.

Would you care to write an implementation that works as an RNNCell? That would be more useful to plug into other experiments (such as character prediction).

Happy to discuss further!

[jiamings]

Hi hardmaru,

Nice to hear from you! I read your blog posts before - they were awesome!

Let me fork your project and check it out on Sunday :)

By the way - I found your comment about the number of the inner loops in
the code. I asked Jimmy Ba - he said he could get away with inner loops of
one, although adding more loops could potentially work better.

Best,
Tony Song
tsong@cs.stanford.edu

[hardmaru]

Cool! I think my implementation is not exactly correct as I had some difficulties understanding the paper. If you want to discuss more later on, feel free to swing by mountain view as well. I'm generally interested in working on these concepts relating to self-modifying weights inside RNNs.

[jiamings]

Sure. I can swing by some time next weekend!

On Sat, Oct 29, 2016 at 3:42 PM hardmaru notifications@github.com wrote:

Cool! I think my implementation is not exactly correct as I had some
difficulties understanding the paper. If you want to discuss more later on,
feel free to swing by mountain view as well. I'm generally interested in
working on these concepts relating to self-modifying weights inside RNNs.

—
You are receiving this because you were assigned.

Reply to this email directly, view it on GitHub
#1 (comment),
or mute the thread
https://github.com/notifications/unsubscribe-auth/AHtfDzO47q0AEb7DYqV_LjjuGFOlERMZks5q48vQgaJpZM4KkNGD
.

[jiamings]

The implementation for RNNCell is here: d98097e

[zhongwen]

Hi, you guys might be interested in this https://github.com/jxwufan/AssociativeRetrival

[hardmaru]

Thanks.