Re: [Scikit-learn-general] scikit-learn @ PyCon Strata 2014

2013-12-13 Thread Mathieu Blondel
On Thu, Dec 12, 2013 at 8:19 PM, Olivier Grisel olivier.gri...@ensta.orgwrote: Also for those who were not aware, Gilles, Andreas and Gael attended NIPS last week and here is the notebook of the presentation Gilles gave at the MLOSS workshop:

[Scikit-learn-general] calibration in sklearn ?

2013-12-13 Thread Eustache DIEMERT
Hi List, I need to test calibration for a given problem, but didn't find anything in sklearn ? If found a pending PR for isotonic calibration [1] but it seems inactive at the moment. Is anyone working on it ? having some gist on one method or another ? Is there interest ? Eustache [1]

Re: [Scikit-learn-general] calibration in sklearn ?

2013-12-13 Thread Olivier Grisel
As far as I know this PR as stalled a bit. I don't know if Alex will have time to work on it in the near future. -- Olivier -- Rapidly troubleshoot problems before they affect your business. Most IT organizations don't

Re: [Scikit-learn-general] calibration in sklearn ?

2013-12-13 Thread Alexandre Gramfort
I don't know if Alex will have time to work on it in the near future. unlikely... sadly... any help very welcome. Alex -- Rapidly troubleshoot problems before they affect your business. Most IT organizations don't

[Scikit-learn-general] SMOTE

2013-12-13 Thread abhishek
hi all, Is there any PR available on SMOTE ? I think it would be good to add SMOTE and ADASYN to scikit-learn and I would like to implement them. If there is any PR available, I would like to start from that. Also, I want to get an opinion of the developers on whether it would be a nice idea to

Re: [Scikit-learn-general] calibration in sklearn ?

2013-12-13 Thread Eustache DIEMERT
Mmmm... if calibration seems to be a good fit for sklearn, I'll try to review the different existing approaches and see if it's difficult to implement the most useful/popular one(s). Any hint on that ? is isotonic regression the most used form or should we have a look on other well-known

Re: [Scikit-learn-general] SMOTE

2013-12-13 Thread Eustache DIEMERT
Do you have any pointer on the relevant literature ? the areas were SMOTE is deemed useful ? E/ 2013/12/13 abhishek abhish...@gmail.com hi all, Is there any PR available on SMOTE ? I think it would be good to add SMOTE and ADASYN to scikit-learn and I would like to implement them. If

Re: [Scikit-learn-general] calibration in sklearn ?

2013-12-13 Thread Olivier Grisel
2013/12/13 Eustache DIEMERT eusta...@diemert.fr: Mmmm... if calibration seems to be a good fit for sklearn, I'll try to review the different existing approaches and see if it's difficult to implement the most useful/popular one(s). Any hint on that ? is isotonic regression the most used form

Re: [Scikit-learn-general] SMOTE

2013-12-13 Thread Dayvid Victor
Hi There, I am working on Instance Reduction (noise removal) on sklearn as my first contribution (hopefully) and I know a SMOTE implementation would be very helpful. Checkout Garcia et all Evolutionary Based IR for imbalanced Datasets. It concludes that SMOTE oversampling is very helpful for

[Scikit-learn-general] Updated KMeansCoder now available as gist

2013-12-13 Thread Kyle Kastner
As a part of some research on dictionary learning, I stumbled across @vene (and others) closed pull request from many moons ago for K-means dictionary learning. I wanted to use it as a testing point, so I updated it for sklearn 0.14 (mixins, some small tweaks to reshaping of stuff in the example,

Re: [Scikit-learn-general] Updated KMeansCoder now available as gist

2013-12-13 Thread Olivier Grisel
Nice. Have you used it with success for real image classification tasks? I see you have been involved in the cats vs dogs kaggle competition. Is learning a linear model, if so we might consider including the such KMeansCoder as part of the sklearn.feature_extraction.image module and write an

Re: [Scikit-learn-general] Updated KMeansCoder now available as gist

2013-12-13 Thread Vlad Niculae
Great, thanks a lot! I'm also curious about what you're running it on and about how the performance is. Vlad On Fri, Dec 13, 2013 at 7:11 PM, Olivier Grisel olivier.gri...@ensta.org wrote: Nice. Have you used it with success for real image classification tasks? I see you have been involved

Re: [Scikit-learn-general] Updated KMeansCoder now available as gist

2013-12-13 Thread Kyle Kastner
I have 2 separate approaches I am considering for real-world testing. For kaggle cats and dogs, using a deep neural network trained on ImageNet (DeCAF http://arxiv.org/abs/1310.1531) for preprocessing, coupled with any kind of classifier, has had excellent success for me so far (even logistic

Re: [Scikit-learn-general] Updated KMeansCoder now available as gist

2013-12-13 Thread Vlad Niculae
We did not implement K-SVD because we did not find any motivation for having two competing dictionary learning implementations, so we stuck with the Julien Mairal et al solver. Do you think that K-SVD would do better than it for this? Vlad On Fri, Dec 13, 2013 at 8:46 PM, Kyle Kastner

Re: [Scikit-learn-general] Updated KMeansCoder now available as gist

2013-12-13 Thread Kyle Kastner
I am not sure that K-SVD is any better, but one of the other people in my group proposed using K-SVD, so I will likely have to implement it. It will be worth a comparison - I will try and stick to the sklearn API. On Dec 13, 2013 2:02 PM, Vlad Niculae zephy...@gmail.com wrote: We did not