can be closed and I can be
rehabilitated in the public eye :)
Best,
Vlad
On Sat, Sep 24, 2011 at 3:25 PM, josef.p...@gmail.com wrote:
On Sat, Sep 24, 2011 at 8:08 AM, Vlad Niculae v...@vene.ro wrote:
The date should be 24th I think since I uploaded it late at night.
You can get it from PyPI
Hopefully the positive reports from the initial discussion are
actually accurate and the new installers are working.
As far as I can tell, sourceforge still hosts the old, buggy
installers. (looking at the date)
If this turns out to be the case, who can update the sourceforge files?
Vlad
On
I think that the binaries without statically linked libgcc and
libstdc++ would have worked if run from a mingw32 environment (ie.
they worked in mine). I wonder if it works the other way around (ie.
no static linking, build with MSVC, run from within mingw32)
Fabian, is the 0.8 release built with
Hello
As far as I can tell (hope I'm not too tired and missing something),
gaussian processes are missing from the class reference. They are not
included in the classes.rst index.
Is this just an omission? Because the module seems to have solid
docstrings that deserve to be listed.
Best,
Vlad
Thank you for the observation. I have been looking into this since
yesterday where the same thing has been reported on my blog by Bob L.
Strum. At the moment I have no idea what the cause is. Does it behave
in the same way if you use the gram solver instead?
Best,
Vlad
On Tue, Oct 18, 2011 at
Interesting. I've been staring at the code but the algorithm itself
shouldn't be losing precision. On the other hand, there are those
stopping conditions that I had taken from the C implementation of
the author of the Cholesky-OMP paper. If it's as you say, it could be
that when it fails, OMP
Hi Jake,
A while back I remember having that issue because my local version of
sphinx was higher than 1.0.0 and thus unsupported by the scikit-learn
docs, so the function links wouldn't work when I built it locally, but
they would work in the online-generated version.
Did the sphinx version used
On Fri, Nov 4, 2011 at 4:54 PM, Andreas Müller amuel...@ais.uni-bonn.de wrote:
On 11/04/2011 03:49 PM, Andreas Müller wrote:
On 11/04/2011 03:42 PM, Alexandre Passos wrote:
On Fri, Nov 4, 2011 at 10:34, Lars Buitinck l.j.buiti...@uva.nl wrote:
2011/11/4 Alexandre Passos alexandre...@gmail.com:
Yes, I was thinking of a sequencial, exploratory IPython-style thing
where you change something in your X and re-fit, when you don't want
to clone and delete the old estimator. Hope this makes sense.
Vlad
2011/11/6 Lars Buitinck l.j.buiti...@uva.nl:
2011/11/6 Vlad Niculae v...@vene.ro
Very much +1, I would always cringe when seeing colormesh calls.
I gave a talk using the IPython notebook and scikit-learn examples,
setting it to SVG mode at the top, and I would have to switch back to
PNG mode for all examples using pcolormesh because it would crash
(well probably just take
Gael, can you give me the info (URL and telephone number) for the
guest house? I would like to call them to make a reservation for me. I
can make a reservation for other people too at the same time if needed
(since Gael only reserved for 5 people).
Yes and maybe we can benefit from Fabian's
On Mon, Nov 21, 2011 at 2:08 PM, Lars Buitinck l.j.buiti...@uva.nl wrote:
2011/11/21 Jacob VanderPlas vanderp...@astro.washington.edu:
I would recommend these: I'm currently taking the Machine Learning
course, taught by Andrew Ng, which will be offered again in January.
It's been a great intro
Hello Massimo
I believe this is an issue others, including me, have faced:
https://github.com/scikit-learn/scikit-learn/issues/445
https://github.com/scikit-learn/scikit-learn/issues/330
I reverted to the stable versions of numpy and scipy from their
website, and the bleeding-edge scikit-learn,
Hi Olivier,
This is very cool. Could we plot average test coverage as well, similar to pep8?
Is there a way to subscribe to the build reports, like with the
buildbot? I signed up but still couldn't find one.
Vlad
On Mon, Nov 28, 2011 at 4:26 AM, Olivier Grisel
olivier.gri...@ensta.org wrote:
On Tue, Nov 29, 2011 at 10:02 PM, Alexandre Gramfort
alexandre.gramf...@inria.fr wrote:
Hi Alex,
I would say:
if it makes sense to fit a GP with only one point:
it should be fixed
Note that even though it might not make any sense in practice, unless
there's a mathematical reason that I'm
On Dec 6, 2011, at 11:04 , Gael Varoquaux wrote:
On Tue, Dec 06, 2011 at 09:41:56AM +0200, Vlad Niculae wrote:
This is actually exactly how the module is designed.
Great design! I should have looked at it closer before writing my mail.
We have BaseDictionaryLearning which only implements
On Tue, Dec 6, 2011 at 11:46 PM, Alexandre Gramfort
alexandre.gramf...@inria.fr wrote:
I do confirm that Lasso and LassoLars both minimize
1/2n || y - Xw || + alpha ||w||_1
and that the n should not be present in the sparse coding context.
it means :
I think I know what happened here.
An upstream change in scipy removed scipy.lena() and left only
scipy.misc.lena().
I wonder if this affects other examples as well. I will try to check
and patch this soon.
Vlad
On Wed, Dec 7, 2011 at 1:38 PM, Gael Varoquaux
gael.varoqu...@normalesup.org
On Dec 8, 2011, at 20:11 , David Warde-Farley wrote:
On Tue, Nov 15, 2011 at 03:13:53AM +0900, Mathieu Blondel wrote:
Hi,
Thanks heaps Gael. I'm planning to contact the guy by tomorrow. I
think it would be easier for him if we don't contact him individually.
I can make the reservations
On Jan 3, 2012, at 17:02 , Olivier Grisel wrote:
2012/1/3 Lars Buitinck l.j.buiti...@uva.nl:
We probably need to extend the sklearn.feature_extraction.text package
to make it more user friendly to work with with pure categorical
features occurrences:
I'm not sure this belongs in
On Jan 5, 2012, at 23:45 , Fabian Pedregosa wrote:
and that was
quite convenient for testing on systems on which nosetest fails
(windows).
Hi Fabian
Could you please be more specific regarding this point, since as a former
Windows user, I find that I don't know what you mean.
On topic, I
Hello all, especially Fabian.
I've noticed that the new examples still don't show up in
scikit-learn.org/dev/, in particular the multi-label one that I'd like to show
off. Can somebody address this?
Sorry if this should be discussed somewhere else.
Best,
Vlad
On Jan 6, 2012, at 17:39 , Olivier Grisel wrote:
2012/1/6 Vlad Niculae zephy...@gmail.com:
Hello all, especially Fabian.
I've noticed that the new examples still don't show up in
scikit-learn.org/dev/, in particular the multi-label one that I'd like to
show off. Can somebody address
Hello everybody,
This is something that has been bugging me for a while. I am not exactly sure
what entity is printing these messages (I assume joblib) but when doing a
verbose CV with n_jobs=1, the progress report looks something like:
[Parallel(n_jobs=1)]: Done job 20 | elapsed: 28.3s
Olivier's solution sounds good.
And it's easy to implement too :) @pprett can you confirm it solves
your perf issue on your data?
I'm talking without actually looking at the code but as long as after fit, the
array will only be needed in F-order, this feels right. However afaik
A bit off topic but since we're talking about work on the SVM module, I noticed
something wrong with the docs.
http://scikit-learn.org/dev/modules/svm.html#tips-on-practical-use
The scaling part makes reference to some Cookbook (I don't know what this is,
it probably died before I joined you
Short answer, no.
sparse_encode is just a wrapper for funcionality that existed in the
scikit already (lasso, omp), with support for parallelization. We
couldn't embed SPAMS anyway, because of the license IIRC.
A benchmark would be interesting indeed.
Vlad
On 09.01.2012, at 18:02, Ian
I am quoting from http://docs.python.org/distutils/builtdist.html
By default the installer will display the cool “Python Powered”
logo when it is run, but you can also supply your own 152x261
bitmap which must be a Windows .bmpfile with the --bitmap option.
I'm assuming -b is short for
On Jan 18, 2012, at 20:23 , Andreas wrote:
On 01/18/2012 07:19 PM, Vlad Niculae wrote:
I am quoting from http://docs.python.org/distutils/builtdist.html
By default the installer will display the cool “Python Powered”
logo when it is run, but you can also supply your own 152x261
bitmap
On Jan 19, 2012, at 00:23 , Gael Varoquaux wrote:
On Wed, Jan 18, 2012 at 07:37:12PM +0900, Mathieu Blondel wrote:
It would be nice if you could make a few contributions to scikit-learn
before the application process starts. This will allow you to
familiarize with the code base, us to
sorry, I don't have a Windows system at the
moment, if you have a VM could you do it? If you're not set up either,
I'll do it in a day or two.
Best,
Vlad
Sent from my iPod
On 27.01.2012, at 12:29, Fabian Pedregosa fabian.pedreg...@inria.fr wrote:
@vene: do you have time to make the windows
My pleasure, I'm sorry for the delay!
Sent from my iPod
On 27.01.2012, at 20:49, Vincent Dubourg vincent.dubo...@gmail.com wrote:
Thank you Vlad! After a slight upgrade of both numpy and scipy I managed
to get a brand new working 0.10 sklearn!
On 27/01/2012 17:32, Gael Varoquaux wrote:
On
Sent from my iPod
On 01.02.2012, at 15:43, Mathieu Blondel math...@mblondel.org wrote:
On Wed, Feb 1, 2012 at 10:10 PM, David Warde-Farley
warde...@iro.umontreal.ca wrote:
I might suggest mean over training examples but sum over output dimensions,
if there is more than one.
Currently,
A nice idea would be to extend the scipy NNLS in the ways needed to use it in
scikit-learn's NMF instead of the _nls_subproblem code translated from C.J.
Lin's code.
The scipy NNLS is written in Fortran. I'd like to bench _nls_subproblem against
it.
Maybe we could have a cython projected sgd
On Feb 3, 2012, at 18:07 , Mathieu Blondel wrote:
On Fri, Feb 3, 2012 at 11:55 PM, Vlad Niculae zephy...@gmail.com wrote:
The scipy NNLS is written in Fortran. I'd like to bench _nls_subproblem
against it.
Maybe we could have a cython projected sgd non-negative least square method
Hi Andre
The installation instructions you are referring to apply only for installing
scikit-learn from source. If you downloaded the binary installer (like you
said) and ran it, there is no need to do `python setup.py install`.
It should work for you to type `import sklearn` in the Python
On Feb 29, 2012, at 21:53 , Olivier Grisel wrote:
2012/2/29 Matthias Ekman matthias.ek...@googlemail.com:
I did some further testing and could reproduce the error on several
machines including a fresh install of debian squeeze using python 2.6.6.
However the problem only occurs with the last
I will try to pick up the work on the one-hot transformer:
https://github.com/scikit-learn/scikit-learn/pull/242
Vlad
On Mar 29, 2012, at 11:36 , Andreas wrote:
Hi Mohit.
Generally all algorithms in sklearn assume that all features are continuous.
Does discrete in your case mean categorial
Hello guys,
Unfortunately I have come down with the flu, and therefore missed a good amount
of time to work on gsoc 2012 proposals. I know that there's not much time left
for review, but here is my pre-proposal for a overall speedup and benchmarking
project.
On Apr 4, 2012, at 17:11 , Olivier Grisel wrote:
Detailed instructions and links on the wiki:
https://github.com/scikit-learn/scikit-learn/wiki/A-list-of-topics-for-a-google-summer-of-code-%28gsoc%29-2012
Please write the draft proposal on a google document or some wiki page
on your
Hi everyone
I have updated my proposal thanks to your excellent suggestions.
I also pointed out the style of optimization that will be applied by linking to
my blog post on optimizing orthogonal matching pursuit code. Unfortunately this
will also flash the bug I introduced before everyone's
Hi David,
Like Gael said in the other thread, try to submit your proposal quite before
the deadline. You can still edit it on their site.
I agree with everybody regarding the importance of testing and examples. They
are not afterthoughts. The documentation, though, can be left until the final
On Apr 6, 2012, at 02:56 , Andreas Mueller wrote:
On 04/05/2012 11:17 PM, Vlad Niculae wrote:
I would like to see a reproduction of the standard neural net digits example:
http://ufldl.stanford.edu/wiki/images/8/84/SelfTaughtFeatures.png
That looks like the weights of an autoencoder
Actually I couldn't find the code but I found something better, the assignment
notes:
https://github.com/SaveTheRbtz/ml-class/blob/master/ex4.pdf
If you ran more iterations it would only get better. Looking back this was a
very good class.
Vlad
On Apr 6, 2012, at 06:54 , Vlad Niculae wrote
On Apr 6, 2012, at 10:19 , Andreas Mueller wrote:
On 04/06/2012 08:04 AM, xinfan meng wrote:
On Fri, Apr 6, 2012 at 1:57 PM, David Warde-Farley
warde...@iro.umontreal.ca wrote:
On 2012-04-05, at 5:17 PM, Vlad Niculae zephy...@gmail.com wrote:
http://ufldl.stanford.edu/wiki/images
Hi Shankar
I am also following the PGM class and I would like to stress out that the way
they implement all the factor operations feels to me to be by no means
efficient, way too much random memory indexing. However the class seems very
insightful, maybe after it ends we will be illuminated as
I think just moving from a train set to a test set would be problematic for
small n_samples.
Vlad
On Apr 17, 2012, at 15:48 , Olivier Grisel wrote:
Le 17 avril 2012 05:39, Gael Varoquaux gael.varoqu...@normalesup.org a
écrit :
On Tue, Apr 17, 2012 at 03:35:26PM +0300, Dimitrios Pritsos
I am very flattered and happy! Thanks to everybody who helped provide this
opportunity.
I think it is very exciting for scikit-learn to have 3 GSoCers, and it's also a
sign of our growth. Congratulations to David and Immanuel, great work so far,
looking forward to interacting as much as we can
For now, lasso (and some others) can be invoked through the sparse_encode
function and it does the multitarget wrapping automatically over multiple
cores. Just pay attention to the shapes of the inputs since they need to be
transposed (the function makes sense in a dictionary learning context).
There has been quite some interest in this in the last couple of months, so I'm
sure it will get some momentum. The question is whether Jake and Olivier's
points about the inappropriateness of the data structures can actually get a
workaround or if this is (more or less) pointless.
If crfsuite
Hello everybody,
I will start my effort for my GSoC project for this year, as discussed, with
making the linear models faster where applicable, most importantly in
multi-task regression problems.
The plan (which will be piloted now, and towards the middle of the summer,
hopefully will get
I can confirm that that exact same test halted for me once too. I thought it
was my old Windows PC that overheated. Sorry for not mentioning it.
Vlad
On May 9, 2012, at 04:11 , Yaroslav Halchenko wrote:
On Wed, 09 May 2012, Olivier Grisel wrote:
so if it fails for some specific seed, I
A significant part of this project will consist of the benchmark suite itself,
that will need to be run by the CI we will deploy.
The question is where to host the benchmark suite. Should I create a new repo
in the scikit-learn project?
scikit-learn/speed
scikit-learn/scikit-learn-speed
On May 11, 2012, at 11:28 , Olivier Grisel wrote:
2012/5/11 Vlad Niculae zephy...@gmail.com:
A significant part of this project will consist of the benchmark suite
itself, that will need to be run by the CI we will deploy.
The question is where to host the benchmark suite. Should I create
On May 28, 2012, at 13:50 , Immanuel B wrote:
Hello,
I could use some feedback on how to best set-up a benchmark for these models:
l2 loss*
log loss*
multi-logit*
with l1 and l1 l2 penalty
Please have a look at the following file:
On May 31, 2012, at 12:42 , Immanuel B wrote:
Does N mean n_samples and p n_features?
yes
What about number of targets, is it 1 everywhere?
not sure what you mean...
The first table contains binary classification data, in the second table the
number of classes is given by #class.
for
This is a consistency question. I found that enet_path has a clever behaviour
for this:
https://github.com/scikit-learn/scikit-learn/tree/master/sklearn/linear_model/coordinate_descent.py#L561
The logic here is:
if center_data changes X, then X wasn't centered.
If this is the case, and the
Congratulations Peter! Excellent work as always!
Vlad
On Jul 5, 2012, at 00:48 , Emanuele Olivetti wrote:
Dear All,
As some of you may have already noticed, Peter (Prettenhofer) has
just won a the Online Product Sales competition on kaggle.com
beating 365 teams:
Hello friends,
As the midterm evaluation is approaching, I pushed the pedal to the metal and
my blog and github profile have seen a lot of activity recently.
I would like to link to everything from one place, and that place will be this
e-mail. So this is what happened:
-- I wrote a couple of
Progress update, I adapted the ml-benchmarks over at
https://github.com/vene/scikit-learn-speed/tree/ml-benchmarks
On Jul 5, 2012, at 15:14 , Olivier Grisel wrote:
Thanks very much for the wrap up Vlad. Could you please document how
to use the %memit and %mrun tools in the performance chapter
Another (hackish) idea to try would be to keep the labels of the extra
data bit give it a sample_weight low enough not to override your good
training data.
On 09.07.2012, at 12:43, Philipp Singer kill...@gmail.com wrote:
Hey!
I am currently doing text classification. I have the following
As per Gael's request, here is my progress compared to what was initially
stated as mid-term goals.
Overall the project is behind schedule, but not far, and I am fairly confident
about its successful completion.
--
GOAL: Set up a running performance benchmark such as speed.pypy.org or Wes
On Jul 11, 2012, at 10:14 , Philipp Singer wrote:
Am 11.07.2012 10:11, schrieb Olivier Grisel:
LinearSVC is based on the liblinear C++ library which AFAIK does not
support sample weight.
Well, that's true.
You should better have a look at SGDClassifier:
This has been merged.
https://github.com/vene/scikit-learn-speed
There are now easy instructions you can follow to run the suite on your own
machine. One step closer to running remotely.
On Jul 6, 2012, at 18:33 , Vlad Niculae wrote:
Progress update, I adapted the ml-benchmarks over
On Jul 12, 2012, at 12:30 , Gael Varoquaux wrote:
On Thu, Jul 12, 2012 at 12:16:50PM +0200, Olivier Grisel wrote:
I get not results...
I haven't followed too much the codebase (I should have but...). That
said, I must confess that I am a bit frightened at the number of
different
On Jul 12, 2012, at 14:10 , Lars Buitinck wrote:
2012/7/12 Gael Varoquaux gael.varoqu...@normalesup.org:
I haven't followed too much the codebase (I should have but...). That
said, I must confess that I am a bit frightened at the number of
different technologies that are being put together.
on
master I can run the suite again and we will have a line with 2 points, yay!
V
On Jul 12, 2012, at 14:25 , Gael Varoquaux wrote:
On Thu, Jul 12, 2012 at 02:22:42PM +0200, Vlad Niculae wrote:
For example a thing that hurts me is that for every 'predict' benchmark, the
model is refitted
, Jul 15, 2012 at 7:07 PM, Vlad Niculae zephy...@gmail.com wrote:
After some bugfixes with Olivier's help, I published the output of the
scikit-learn-speed here:
http://vene.github.com/scikit-learn-speed/
Because there is only one data point, it looks like the plots are empty, but
you can
A preview of the HTML benchmarking report that will soon be deployed:
http://blog.vene.ro/2012/07/20/scikit-learn-speed-html-reports-teaser/
Best,
Vlad
On Jul 16, 2012, at 20:05 , Peter Prettenhofer wrote:
It seems like vbench is failing when it tries to run the following git
command::
!
The tests just need some cleaning up, and scikit-learn-speed will soon be up
and running!
On Jul 21, 2012, at 15:54 , Vlad Niculae zephy...@gmail.com wrote:
A preview of the HTML benchmarking report that will soon be deployed:
http://blog.vene.ro/2012/07/20/scikit-learn-speed-html-reports
Either way, is there a reason that I'm missing, why np.array([0]) should be
both C- and F-contiguous, but np.array([[0]]) can only be one of them at a time?
On Aug 2, 2012, at 17:26 , Olivier Grisel olivier.gri...@ensta.org wrote:
2012/8/2 Skipper Seabold jsseab...@gmail.com:
On Thu, Aug 2,
A full benchmark suite has been run successfully:
http://jenkins-scikit-learn.github.com/scikit-learn-speed/
The whole process (build sklearn, run benchmarks, generate output) took ~40
minutes, so I am scheduling it once a week.
Best,
Vlad
On Jul 29, 2012, at 20:49 , Vlad Niculae zephy
Andy, Mathieu:
The docs are lacking guidelines and examples on how to tune SVR parameters.
IIUC, C, gamma, etc should be use just as in SVC. The tricky part is epsilon,
how should it be set? What are sensible defaults and a sensible grid search
range?
Thanks,
Vlad
On Aug 9, 2012, at 13:30 ,
On Aug 16, 2012, at 18:57 , iBayer mane.d...@googlemail.com wrote:
Hi,
I know it sounds stupid but where is the code for ``as_float_array``?
because of:
It's in `validation.py`, you can find this out either by looking where it's
imported from in the `__init__.py` or by using the
-learn-speed). We can then link to this from the
homepage.
What do you think?
Best,
Vlad
On Aug 6, 2012, at 11:26 , Vlad Niculae zephy...@gmail.com wrote:
A full benchmark suite has been run successfully:
http://jenkins-scikit-learn.github.com/scikit-learn-speed/
The whole process (build
On Aug 17, 2012, at 21:10 , Olivier Grisel olivier.gri...@ensta.org wrote:
2012/8/17 Vlad Niculae zephy...@gmail.com:
If the build scheduled tonight runs successfully, with the newly added
benchmarks, I would like to move the scikit-learn-speed codebase to
scikit-learn/scikit-learn-speed
On Aug 19, 2012, at 14:16 , Vlad Niculae zephy...@gmail.com wrote:
On Aug 17, 2012, at 21:10 , Olivier Grisel olivier.gri...@ensta.org wrote:
2012/8/17 Vlad Niculae zephy...@gmail.com:
If the build scheduled tonight runs successfully, with the newly added
benchmarks, I would like
This is confusing to me too. I wanted to copy it for the OMP CV, but it seems
overly complicated.
Vlad
On Aug 20, 2012, at 17:19 , Andreas Müller amuel...@ais.uni-bonn.de wrote:
Hi Alex.
Thanks for the answer. So it estimates ``n_nonzero_coefs``.
As far as I can see, you can not get this
On Aug 20, 2012, at 04:06 , Vlad Niculae zephy...@gmail.com wrote:
On Aug 19, 2012, at 14:16 , Vlad Niculae zephy...@gmail.com wrote:
On Aug 17, 2012, at 21:10 , Olivier Grisel olivier.gri...@ensta.org wrote:
2012/8/17 Vlad Niculae zephy...@gmail.com:
If the build scheduled tonight
On Aug 22, 2012, at 20:38 , Olivier Grisel olivier.gri...@ensta.org wrote:
FYI and as a side note for this GSoC project: I have just filled in
the final evaluation form for Vlad's GSoC project and gave it a pass:
congrats Vlad :)
\o/ Thank you Olivier, this has been a very enjoyable project
We are all annoyed by warnings; we have a ton of them at the moment. Some of
them are scheduled for removal, and others have even passed their deadline.
I think we should go through them thoroughly before the release. I could
volunteer for this.
Best,
Vlad
--
Vlad N.
On Aug 31, 2012, at 18:27 , amuel...@ais.uni-bonn.de wrote:
We do? Which warnings do you mean? I am not aware of any warnings in the
tests or examples.
Sorry, I exaggerated because I was looking at the latest release instead. The
test suite is clean, but the codebase still has some leftover
On Sep 6, 2012, at 18:08 , Mathieu Blondel math...@mblondel.org wrote:
Hello,
The Perceptron can be seen as a SGD algorithm optimizing the loss \sum_i
max{t - y_i w^T x_i, 0} where t=0. On the other hand, online SVM optimizes
the same loss but with t=1 (the advantage of setting t=1
On Oct 1, 2012, at 11:22 , Alexandre Gramfort alexandre.gramf...@inria.fr
wrote:
That's great news! Is this connected to your image processing seminar?
It's related to my new position at ParisTech but image processing and
ML are taught in different classes.
That's great, congratulations
Hi Jaidev,
This seems relevant to your question:
http://metaoptimize.com/qa/questions/7897/are-lasso-and-basis-pursuit-really-the-same-thing
Vlad
On Oct 5, 2012, at 01:16 , Jaidev Deshpande deshpande.jai...@gmail.com wrote:
Hi,
I've been going through the tomography reconstruction example
Then, on Sunday, we can work on the release itself (building binaries,
uploading the webpage...).
How does that sound?
Thanks Brian for volunteering to help with the Windows binaries. In case your
schedule is tight I can step in too.
Vlad
If you have the time to do the cherry picking,
If Brian is not available tomorrow morning/afternoon I can build and upload the
win32 binaries. I didn't manage to set up win64 virtualenvs though.
Vlad
On Oct 7, 2012, at 21:58 , Gael Varoquaux gael.varoqu...@normalesup.org wrote:
Sorry guys, I've had loads of stuff on and I might have a
I'm on the Win32 binaries.
Vlad
On Oct 9, 2012, at 00:14 , Gael Varoquaux gael.varoqu...@normalesup.org wrote:
On Tue, Oct 09, 2012 at 12:27:54AM +0200, Gael Varoquaux wrote:
I can't really give instruction, as I have never shipped windows
binaries. The core idea is to run 'python setup.py
On Oct 9, 2012, at 09:17 , Andreas Mueller amuel...@ais.uni-bonn.de wrote:
Thanks Gael for pulling of the release basically single-handedly :)
And now, off we go towards 0.13!
And/or 1.0!
Vlad
--
Don't let slow site
More importantly, the process of building Windows binaries should not
need make, or anything else outside of what `python setup.py` can do.
My 2c,
Vlad
On Mon, Oct 15, 2012 at 7:59 AM, Gael Varoquaux
gael.varoqu...@normalesup.org wrote:
On Mon, Oct 15, 2012 at 07:49:23AM +0100, Brian Holt
Do you think we should add a check so the concatenation is not
performed if the docstring is None? The increased binary compression
might matter enough for some users.
Vlad
On Wed, Oct 17, 2012 at 2:13 PM, Legault, Alain
alain.lega...@rncan-nrcan.gc.ca wrote:
Brent hit it right on!
I had
Also, since you're using scikit-learn, you could try giving the joblib `dump`
and `load` a go. Joblib is bundled with scikit-learn: `from
sklearn.externals.joblib import dump, load`. They support various degrees of
compression and were designed for saving large models.
Vlad
On Wed, Oct 17,
On Oct 18, 2012, at 16:58 , Gael Varoquaux gael.varoqu...@normalesup.org
wrote:
On Thu, Oct 18, 2012 at 05:56:23PM +0200, Olivier Grisel wrote:
Even though it's not officially supported by Apple, the bug seems to
have been fixed in 10.8.
Awesome, that's good news!
It's not completely
Hello,
It seems I have reached again the need for something that became
apparent when working with image patches last summer. Sometimes we
don't have a 1 to 1 correspondence between samples (rows in X) and
actual documents we are interested in scoring over. Instead, each
document consists of (a
and let it pass throught when they do their
job, so the grouping can be fed in with the dataset and used at the end during
scoring.
The text feature extraction sort of deals with this by using a list, right?
I'm not sure what you mean by this.
Cheers,
Andy
On 10/31/2012 01:13 PM, Vlad Niculae
On Nov 14, 2012, at 14:09 , Olivier Grisel olivier.gri...@ensta.org wrote:
I would have also liked to implement a hashing text vectorizer but I
am not sure I will find the time to do this week or the next week.
I'd love to help with that next week!
--
Vlad N.
http://vene.ro
In the matrix-matrix case (as opposed to vector-vector or matrix-vector), I
played with Mathieu's dot-bench and it didn't beat Scipy's very efficient
implementation.
On Fri, Dec 28, 2012 at 7:51 AM, Mathieu Blondel math...@mblondel.orgwrote:
I forgot to mention that the multiplication of two
Maybe my mind is not in its right place but how is that different from
using the PCA transformer?
On Thu, Jan 3, 2013 at 10:48 PM, Lars Buitinck l.j.buiti...@uva.nl wrote:
2013/1/3 Jack Alan j.o.alan2...@gmail.com:
I'm working in document classification and I wonder if there is a way of
PR #804 had some comments about generating the tables automatically, which
would be nice. How about a consistently structured `Complexity` section to
the docstrings, and use it to populate the table?
On Thu, Jan 10, 2013 at 6:38 PM, Ronnie Ghose ronnie.gh...@gmail.comwrote:
yes please. I was
Olivier, the histogram plotting and data transformation is great, valuable
practical advice that would be nice to have in the docs. I think it would
go nicely as part of a tutorial, what do you think?
Vlad
On Fri, Jan 11, 2013 at 10:38 AM, Olivier Grisel
olivier.gri...@ensta.orgwrote:
1 - 100 of 233 matches
Mail list logo