Le 29 mars 2012 23:55, Alexandre Passos <alexandre...@gmail.com> a écrit : > On Thu, Mar 29, 2012 at 17:43, Daniel Duckworth <duckwor...@berkeley.edu> > wrote: >> Has someone taken ownership of sklearn.hmm? If the task is open or help >> would be appreciated, I'd like to jump in. > > That's great. I know I have no idea why it has some of the bugs it > has, which I tripped all over while trying to implement the DPGMM.
I have submitted a PR on simplifying the test hierarchy to make it possible to support Python 3 in the future: https://github.com/scikit-learn/scikit-learn/pull/748 Any review appreciated. As for the owneer ship of this module as you can see there are XXX and SkipExceptions in those tests. In particular they check that the training likelihood should always increase but this is not always the case for some models: playing with the RNG seed can break those tests besides the fact that the tolerance (-0.5) seems already high to me. If someone want to investigate that would be great. -- Olivier http://twitter.com/ogrisel - http://github.com/ogrisel ------------------------------------------------------------------------------ This SF email is sponsosred by: Try Windows Azure free for 90 days Click Here http://p.sf.net/sfu/sfd2d-msazure _______________________________________________ Scikit-learn-general mailing list Scikit-learn-general@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
