2012/8/1 Andreas Müller <[email protected]>:
> Hey Everybody.
> Soon it is time for a new release. I notices some people are going on holiday 
> (I think Lars and Gilles),
> so maybe it would be good for wait them to be back.

I will be off the next 2 weeks. The weekend after that I will attend
EuroScipy in Brussels so I won't be only for helping with the release
either.

> Therefore I'd suggest releasing at the end of August.
> Any opinions?

In mid-september there will be a scikit-learn sprint in Paris:
http://www.pycon.fr/2012/ but we don't need to wait for that sprint to
make a new release.

> My wish list looks like this (from more realistic to less realistic):
> - Precomputed and Callable kernels for sparse SVM #920
> - what used to be the fit_pairwise proposal, now gives more unified API for 
> clustering #803
> - Single codebase Python 3 #702
> - Multinomial Logistic Regression SGD #849
>
> As well as many issues that need be investigated.
> Any help on that front would be greatly appreciated!

I don't want to push any particular feature / fix for that release (or
any other release BTW). I think it's best to release what has been
merged to master every 3-4 months and let the regular review work
happen without any time pressure.

But it's great to make regular release, so definitely +1 for getting
something out by the end of the summer!

FYI my personal goals in the short to medium term for scikit-learn is
to make it easier to work with semi-largish datasets without
exhausting the memory on a large amazon machine with 8 to 16
concurrent threads (either physical cores or hyper threading) and
~64GB of RAM. I think it's important to adress that use case as I
think it's the typical environment for solving tasks such as kaggle
challenges that got the project some recent advertisement and new
users in particular thanks to Peter and Emanuele recent successes in
kaggle competitions.

So for instance I would like to help fix the following recently reported issues:

- https://github.com/scikit-learn/scikit-learn/issues/325 (old but
relevant to the KNN memory bug reported on the ML)
- https://github.com/scikit-learn/scikit-learn/issues/936 (random
forest memory usage when n_jobs != 1)

-- 
Olivier
http://twitter.com/ogrisel - http://github.com/ogrisel

------------------------------------------------------------------------------
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

Reply via email to