Re: [Scikit-learn-general] Adding new classifiers

2013-07-25 Thread Andreas Mueller
Hi Michal. The documentation is here: http://scikit-learn.org/dev/developers/index.html#rolling-your-own-estimator It is indeed a tad hard to find currently. We are working on a slight reorganization of the docs ;) If you are ok with relying on scikit-learn, you can inherit from BaseEstimator and

[Scikit-learn-general] Adding new classifiers

2013-07-25 Thread Michal Romaniuk
Hi, I'm working on a customized classifier and I would like it to be compatible with sklearn, so that I can use it with pipelines, GridSearchCV and replicate it using sklearn's clone function. I've looked at the code for some classifiers but I'm not sure which base classes to use. Is there any doc

Re: [Scikit-learn-general] Beta / candidate release this weekend

2013-07-25 Thread Lars Buitinck
2013/7/25 Jacob Vanderplas : > On Thu, Jul 25, 2013 at 7:44 AM, Andreas Mueller > wrote: >> >> There seems to be also a separate issue about cython compatibility. The >> current cython version >> doesn't like sklearn/neighbors/binary_tree.pxi. It warns about undefined >> behavior because you use i

Re: [Scikit-learn-general] Beta / candidate release this weekend

2013-07-25 Thread Jacob Vanderplas
On Thu, Jul 25, 2013 at 7:44 AM, Andreas Mueller wrote: > > There seems to be also a separate issue about cython compatibility. The > current cython version > doesn't like sklearn/neighbors/binary_tree.pxi. It warns about undefined > behavior because you use indexing like X[:-1] > when running cyt

Re: [Scikit-learn-general] Beta / candidate release this weekend

2013-07-25 Thread Andreas Mueller
On 07/25/2013 03:22 PM, Jacob Vanderplas wrote: > There's still the numpy 1.4 incompatibility issue in the new ball tree > code [1]... I've been absolutely swamped this week and haven't had a > chance to address the problem yet. I'll see if I can look into it > today or tomorrow. I'm still not

Re: [Scikit-learn-general] GridSearchCV with multi-label: ROC-AUC-equivalent metrics

2013-07-25 Thread Josh Wasserstein
Thank you Andreas. That's what I thought, which is why I asked this question in StackOverflow, but apparently --upgrade does not work with git repositories. I think uninstalling and re-installing again will

Re: [Scikit-learn-general] Beta / candidate release this weekend

2013-07-25 Thread Jacob Vanderplas
There's still the numpy 1.4 incompatibility issue in the new ball tree code [1]... I've been absolutely swamped this week and haven't had a chance to address the problem yet. I'll see if I can look into it today or tomorrow. I'm still not sure about the best approach to fixing this, so it may tak

Re: [Scikit-learn-general] GridSearchCV with multi-label: ROC-AUC-equivalent metrics

2013-07-25 Thread Andreas Mueller
On 07/25/2013 02:53 PM, Josh Wasserstein wrote: Thanks Joel. Before posting my post I did the following, so I don't know why I didn't get the very latest version: $ pip install git+git://github.com/scikit-learn/scikit-learn@master Download

Re: [Scikit-learn-general] GridSearchCV with multi-label: ROC-AUC-equivalent metrics

2013-07-25 Thread Josh Wasserstein
Thanks Joel. Before posting my post I did the following, so I don't know why I didn't get the very latest version: $ pip install git+git://github.com/scikit-learn/scikit-learn@master Downloading/unpacking git+git://github.com/scikit-learn/scikit-learn@master Cloning git://github.com/scikit-lear

Re: [Scikit-learn-general] GridSearchCV with multi-label: ROC-AUC-equivalent metrics

2013-07-25 Thread Joel Nothman
On the contrary, make_scorer, replacing Scorer, was merged into master in the last couple of days. Try pulling the latest changes. On Thu, Jul 25, 2013 at 10:33 PM, Josh Wasserstein wrote: > Got it, I just realized that the > dev

Re: [Scikit-learn-general] GridSearchCV with multi-label: ROC-AUC-equivalent metrics

2013-07-25 Thread Josh Wasserstein
Got it, I just realized that the devdocumentation is outdated (looking at the code I noticed that make_scorer has been replaced by Scorer). Thanks. Josh On Thu, Jul 25, 2013 at 8:24

Re: [Scikit-learn-general] GridSearchCV with multi-label: ROC-AUC-equivalent metrics

2013-07-25 Thread Josh Wasserstein
Thanks. I am having problems when using the micro/macro variants for GridSearchCV. I tried creating the corresponding scorer objects, but I got the error: > cannot import name make_scorer This is with 0.14 git (from master) that I checked out about a week ago. Here is the code in more detail

Re: [Scikit-learn-general] GridSearchCV with multi-label: ROC-AUC-equivalent metrics

2013-07-25 Thread Olivier Grisel
2013/7/25 Josh Wasserstein : > Thank you Olivier. I went through that paper and I agree, it looks like > implementing micro-AUC or macro-AUC should not be that hard. I will try to > implement within the next week. I have have never contributed to a project > in GitHub, so I am not sure to what ext

Re: [Scikit-learn-general] GridSearchCV with multi-label: ROC-AUC-equivalent metrics

2013-07-25 Thread Josh Wasserstein
Thank you Olivier. I went through that paper and I agree, it looks like implementing micro-AUC or macro-AUC should not be that hard. I will try to implement within the next week. I have have never contributed to a project in GitHub, so I am not sure to what extent my code would meet the standards

[Scikit-learn-general] Beta / candidate release this weekend

2013-07-25 Thread Andreas Mueller
Hey everybody. We plan to make a beta release / release candidate this weekend. There will be a feature freeze on Friday for everything except bug fixes.( I'm now trying to tag stuff that I think is relevant: https://github.com/scikit-learn/scikit-learn/issues?milestone=8 Feedback very welcome. W