On 07/23/2012 11:49 AM, Peter Prettenhofer wrote: > 2012/7/23 Emanuele Olivetti <[email protected]>: >> [..] >> >> As you can see the solution is very simple and just based on >> np.logaddexp.reduce() >> instead of np.exp().sum(), plus np.nan_to_num() and a little rearrangement. >> I can prepare a pull request if you are interested. > Hi Emanuele, > > Absolutely, that would be great - please, also include a test which > exposes the bug. > >> With this little patch the issue disappears and GradientBoostingClassifier >> gives the expected answers. >> >> Note that I haven't a simple toy example to reproduce the issue and >> the actual dataset I'm using is large. Anyway I am sure that with a >> little bit of time it would be possibile to write a simple example >> that tricks MultinomialDeviance as described above. > I see - anyways, it would be great if you could prepare the patch in a > PR and we can work on the test collaboratively. > > Lets keep Scott posted - he might be interested as well @scottblanc . > > thanks, > Peter > > >
Hi, Pull request #975 , https://github.com/scikit-learn/scikit-learn/pull/975 is ready for comments and contributions. It just incorporates the fix I proposed above with logsumexp() instead of logaddexp.reduce(), following the suggestion of Gael. The PR lacks a proper simple test. I'll try to think about it. Nevertheless help is much appreciated :) Best, Emanuele ------------------------------------------------------------------------------ Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ _______________________________________________ Scikit-learn-general mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
