Re: [Scikit-learn-general] Custom Scorer

> This patch doesn't extend scorers, but it's not hard to make your own. This statement of mine was nonsense :) It certainly extends scorers to handle `sample_weight`. On 7 January 2014 13:49, Joel Nothman wrote: > Hi Mark, > > Noel Dawe has implemented support for weighted samples in a number

Re: [Scikit-learn-general] Custom Scorer

Hi Mark, Noel Dawe has implemented support for weighted samples in a number of metrics and in grid search / cross validation, but it is not yet merged: https://github.com/scikit-learn/scikit-learn/pull/1574. It currently does not includ

Re: [Scikit-learn-general] Custom Scorer

Thanks Andy, Would GridSearchCV and cross_val_score be usable for the following? I want to report on a classifier's in relation to some form of cost function. Eg. Intervention programs show there is a $ cost of a FP and a $ gain of a TP (which change based on characteristics of the churn event).

Re: [Scikit-learn-general] Save trained classifier

Hello, I have a trained ExtraTreesRegressor saved using joblib.dump (without compress). This creates more than ten thousand files, each weighing less than 100KB. When trying to load using joblib.load with mmap_mode="r" I get a [Errno 24] Too many open files. Is there a way to save my regressor wit

Re: [Scikit-learn-general] Custom Scorer

Hi Mark. Currently there is no good way to do this when using GridSearchCV or cross_val_score (which I guess is your question). I think the easiest way is to write the loop over folds yourself using IterGrid. We are aware of the use case but didn't really have the man-power to investigate it.

[Scikit-learn-general] Custom Scorer

Hi all, I'm interested in using several of the standard classification metrics (precision_recal_curve, roc_curve, classification_report) but weighted based on another variable associated with the prediction. In concrete terms, I'm predicting churn and I want each of the above classification metri

[Scikit-learn-general] Dev website again

Hey everybody. I noticed the dev website wasn't build for a while now. Is anyone on that? Cheers, Andy -- Rapidly troubleshoot problems before they affect your business. Most IT organizations don't have a clear picture

Re: [Scikit-learn-general] Possible replacement for Gaussian Process module

On 01/05/2014 08:09 AM, Alexandre Gramfort wrote: > hi, > > I feel the same as Jake. +1 > - we start a deprecation cycle introducing new GaussianProcessRegressor > and GaussianProcessClassifier objects to avoid too much of a mess > > - we advertise this code as sklearn third party on the wiki. I h

Re: [Scikit-learn-general] Bumping the dependencies (numpy and scipy) to the versions from Ubuntu Precise 12.04 LTS

+1 On Jan 6, 2014 1:35 PM, "Gael Varoquaux" wrote: On Mon, Jan 06, 2014 at 05:25:33AM -0800, Jacob Vanderplas wrote: > Is there a reason not to bump Scipy up to 0.10? I'd rather be on the latest LTS. Many insitutions need to be on an LTS, and having to install packages outside of the standard sy

Re: [Scikit-learn-general] Bumping the dependencies (numpy and scipy) to the versions from Ubuntu Precise 12.04 LTS

On Mon, Jan 06, 2014 at 05:25:33AM -0800, Jacob Vanderplas wrote: > Is there a reason not to bump Scipy up to 0.10? I'd rather be on the latest LTS. Many insitutions need to be on an LTS, and having to install packages outside of the standard system is a big increase in sysadmin burden on everybo

Re: [Scikit-learn-general] Bumping the dependencies (numpy and scipy) to the versions from Ubuntu Precise 12.04 LTS

On Mon, Jan 6, 2014 at 4:23 AM, Olivier Grisel wrote: > > > - numpy 1.6.1+ (see http://packages.ubuntu.com/precise/python-numpy) > - scipy 0.9.0+ (see http://packages.ubuntu.com/precise/python-scipy ) > +1 as well: especially with build systems like conda and wheels becoming more widespread, I don

Re: [Scikit-learn-general] Bumping the dependencies (numpy and scipy) to the versions from Ubuntu Precise 12.04 LTS

2014/1/6 Andy : > On 01/06/2014 12:24 PM, Gael Varoquaux wrote: >> On Mon, Jan 06, 2014 at 01:23:06PM +0100, Olivier Grisel wrote: >>> Therefore I think it is time to bump the minimal versions for our >>> dependencies to the versions shipped with Ubuntu Precise 12.04 LTS: >>> - numpy 1.6.1+ (see ht

Re: [Scikit-learn-general] Bumping the dependencies (numpy and scipy) to the versions from Ubuntu Precise 12.04 LTS

On 01/06/2014 12:24 PM, Gael Varoquaux wrote: > On Mon, Jan 06, 2014 at 01:23:06PM +0100, Olivier Grisel wrote: >> Therefore I think it is time to bump the minimal versions for our >> dependencies to the versions shipped with Ubuntu Precise 12.04 LTS: >> - numpy 1.6.1+ (see http://packages.ubuntu.c

Re: [Scikit-learn-general] Bumping the dependencies (numpy and scipy) to the versions from Ubuntu Precise 12.04 LTS

On Mon, Jan 06, 2014 at 01:23:06PM +0100, Olivier Grisel wrote: > Therefore I think it is time to bump the minimal versions for our > dependencies to the versions shipped with Ubuntu Precise 12.04 LTS: > - numpy 1.6.1+ (see http://packages.ubuntu.com/precise/python-numpy) > - scipy 0.9.0+ (see htt

[Scikit-learn-general] Bumping the dependencies (numpy and scipy) to the versions from Ubuntu Precise 12.04 LTS

Hi all, We had this discussion several times in pull-request's discussion threads but AFAIK never on the mailing list: It has become an increasing burden to maintain compatibility with very old numpy and scipy versions. In particular I recently broke the jenkins 2.6 job by introducing a change th

Re: [Scikit-learn-general] [Job] Data Science Consultant

Hey Thorsten. Too bad you didn't give me a shout when Bonn was the German hub for sklearn development ;) I didn't know about your company. I'll tell my friends. Cheers, Andy On 01/06/2014 10:08 AM, Thorsten Kranz wrote: Good morning, I've been using scikit-learn since its early days. Now m

Re: [Scikit-learn-general] Job proposal - welcome or off-topic?

Thank you, Gael and Olivier. Olivier, good point, I'll stick to this habbit. Greetings, Thorsten 2014/1/6 Olivier Grisel > +1 but maybe we could require the posters to use a "[JOB]" object > prefix to explicitly mark such ads and make it possible for > subscribers to filter them. > > -- > Oli

[Scikit-learn-general] [Job] Data Science Consultant

Good morning, I've been using scikit-learn since its early days. Now my company is seeking a Python expert as Data Science Consultant. The company is located in Bonn, Germany, and supplies big companies (mainly insurance and automotive industry) with software for Business Intelligence and Data An

Re: [Scikit-learn-general] Job proposal - welcome or off-topic?

+1 but maybe we could require the posters to use a "[JOB]" object prefix to explicitly mark such ads and make it possible for subscribers to filter them. -- Olivier -- Rapidly troubleshoot problems before they affect you

Re: [Scikit-learn-general] Job proposal - welcome or off-topic?

On Mon, Jan 06, 2014 at 10:41:50AM +0100, Thorsten Kranz wrote: > before polluting this list - are job proposals, if related to machine > learning, welcome on this list or would you regard this as spam? If they are really job proposals, and have a scikit-learn focus, I wouldn't regard them as spam

[Scikit-learn-general] Job proposal - welcome or off-topic?

Hi, before polluting this list - are job proposals, if related to machine learning, welcome on this list or would you regard this as spam? Greetings, Thorsten -- Rapidly troubleshoot problems before they affect your busin