Closed Saurabh7 closed 10 years ago
LGTM :smile:
Wow this somehow slipped through my fingers so far. Checking it out tomorrow
@karlnapf thanks! I guess you weren't around on irc when i talked about this...
@sonney2k Could you have a look here. I have no objections to merge this, but I have no idea how to test this.
@Saurabh7 oh sorry I totally missed this. Looking at it now.
@sonney2k any updates here?
@karlnapf hi, some changes were suggested, i will push them today... but i think I can add something more like you suggested before, model-selection maybe ?since selecting likelihood model is already added
Model selection for all classification demos would be nice in fact. See the GP notebook for how to do that. @kislay is also on doing the same thing for GP regression demo
Added an option to do model selection
Nice - code looks good already Sorry about the massive delay here. I will download very soon and try locally. And then send some better feedback
@vigsterkr @sonney2k it would be amazing if we could somehow run the demos before merge, same for notebooks, its a pain to do this manually
Does this fully work locally on your machine with latest Shogun version?
Yup, this works with latest shogun locally :)
Ok then, I will merge now to see what happens. Pls send a PR with the minor comments addressed Looking forward to play with that.
Ok, for some proper feedback!
This is too slow, could you reduce the grid-size to make it run faster?
The decision boundary should be bold, to clearly stick out the heat map.
Please add parameters for kernel scale and noise (all the GP inference parameters)
When running locally, the dataset importer does not work. Could you either add this to the howto, or fix it?
I get
import h5py, json
ImportError: No module named h5py
This should be mentioned in the howto, as well as the django requirements
I just installed and now it works - but the dataset makes about zero sense for binary classification. Why is this one in there?
Also, it is so slow that running the GP on the dataset takes more than a few minutes. Not good!
Hi, I added a demo for gp classification as suggested in http://www.shogun-toolbox.org/page/Events/gsoc2014_ideas#missionary. I tested it locally and it worked fine for me :)