uRNN code

[casperkaae]

Hi Amar and Martin

I've read your uRNN paper and the following reddit discussion with great interest. At reddit you mention that you have a code that is 4x faster than the one on github.
https://www.reddit.com/r/MachineLearning/comments/3uk2q5/151106464_unitary_evolution_recurrent_neural/cxfwsqw

is the current github version the faster code or is that not released yet?

BR Casper

[martinarjovsky]

It is indeed the current version on github.

Best,
Martin
On Jan 5, 2016 10:38 AM, "Casper Sønderby" notifications@github.com wrote:

Hi Amar and Martin

I've read your uRNN paper and the following reddit discussion with great
interest. At reddit you mention that you have a code that is 4x faster than
the one on github.

https://www.reddit.com/r/MachineLearning/comments/3uk2q5/151106464_unitary_evolution_recurrent_neural/cxfwsqw

is the current github version the faster code or is that not released yet?

BR Casper

—
Reply to this email directly or view it on GitHub
#1.

[casperkaae]

Great, thanks

The results seems very impressive! Any chance you have some results from experiments on other datasets that you would like to share ?

[martinarjovsky]

Thanks! We do have similar results on other datasets, but they are still a
work in progress as we are trying different new versions of the model. We
will publish them in the future when they are ready.

As a side note to the original comment, we will continue to push
optimization changes as well since there is still room for improvement.

Best!
Martin
On Jan 5, 2016 10:53 AM, "Casper Sønderby" notifications@github.com wrote:

Great, thanks

The results seems very impressive! Any chance you have some results from
experiments on other datasets that you would like to share ?

—
Reply to this email directly or view it on GitHub
#1 (comment).

[casperkaae]

Thats good to hear, sometimes it's struggle when leaving MNIST :)

Thanks for sharing the code

BR
Casper