search

dicarlolab / archconvnets

Architecturally optimized neural networks trained with regularized backpropagation
5 stars 5 forks source link

Modified code to allow caching, added module for loading from a referenc... #22

Closed ardila closed 10 years ago

ardila commented 10 years ago

...e model

ardila commented 10 years ago

@yamins81 can you take a look so that my branches don't get too far behind?

yamins81 commented 10 years ago

Having a look know. I see how this code handles the issue of boolean labels (that's good). But where is the stuff for leading from a reference model? And why specifically do we need that?

ardila commented 10 years ago

It's for loading pieces of models, which is something we both mentioned potentially needing. I made it when I was working on fitting to neurons.

On Thu, May 1, 2014 at 8:58 PM, Dan Yamins notifications@github.com wrote:

Having a look know. I see how this code handles the issue of boolean labels (that's good). But where is the stuff for leading from a reference model? And why specifically do we need that?

— Reply to this email directly or view it on GitHubhttps://github.com/dicarlolab/archconvnets/pull/22#issuecomment-41973646 .

yamins81 commented 10 years ago

So it's not particularly about reference models per se. Also I don't see the code for this, though it's mentioned in the commit message.

ardila commented 10 years ago

I think that's in the other pull request, and won't be merged unless you merge that one as well