Daniel Nouri
Daniel Nouri
FWIW, Krizhevsky 2012 has a plot with the filters learned in the first conv layer.
The PR "Add ability to weight training samples and predefine training/validation splits" #123 has an implementation for this. The pull request has problems though which are discussed here: https://github.com/dnouri/nolearn/pull/123#issuecomment-124687911
The best way to go about this might be to create a second input layer for the weights, then use that layer inside of your `objective` function to multiply your...
Hmm, would that be possible by creating a custom `objective` function alone?
What was the argument again for not matching by names? I think that would be more transparent and robust. (Also, please squash your commits, here's a guide: http://gitready.com/advanced/2009/02/10/squashing-commits-with-rebase.html)
(please squash this time, it's easy) Not sure why you think matching by names wouldn't be done automatically? Say in one case, you might want to start out with a...
On Fr, Apr 24, 2015 at 2:30 , Benjamin Bossan [email protected] wrote: > Okay, I misunderstood what you meant, I thought the user would have > to spell out which...
Good call. I'll try and start this effort over the next week or two. Thanks for offering your help.
Accidental close
So I assume what you're doing right now to achieve this is set an attribute on your custom batch iterator object after you instantiate `NeuralNet`, maybe like so: ``` python...