2012/1/19 ert
> I wanted to train a multinomialnb classifier on a training set containing
> 18k features each containing about 1.8M examples .. Unfortunately I do not
> have enough memory ( I have 4G) on my system to create such an array...Is
> there a sparse implementation of Naive Bayes which c
Hi,
I wanted to train a multinomialnb classifier on a training set containing 18k
features each containing about 1.8M examples .. Unfortunately I do not have
enough memory ( I have 4G) on my system to create such an array...Is there a
sparse implementation of Naive Bayes which can be used? or i
2012/1/19 Kenneth C. Arnold :
> As an aside to those who use scipy's sparse matrices: do you find it
> troublesome that scipy's sparse things behave like matrices instead of
> like ndarrays? If dense matrices are a thin wrapper around dense
> ndarrays, shouldn't sparse matrices be a thin wrapper ar
2012/1/19 Kenneth C. Arnold :
> Divisi[1] uses PySparse[2,3].
>
> [1] https://github.com/commonsense/divisi2
> [2] http://pysparse.sourceforge.net/
> [3] https://github.com/rspeer/csc-pysparse
No first hand experience here, but I believe the (deprecated)
scipy.maxent supports that as well.
--
La
On Thu, Jan 19, 2012 at 11:03 AM, Satrajit Ghosh wrote:
> in one of my projects i use the scipy sparse library for turning a graph
> into a sparse dependency matrix and then manipulating this matrix
> (adding/subtracting columns/rows, setting elements to 0, ...). this is the
> only reason i have s
hi all,
in one of my projects i use the scipy sparse library for turning a graph
into a sparse dependency matrix and then manipulating this matrix
(adding/subtracting columns/rows, setting elements to 0, ...). this is the
only reason i have scipy as a dependency and would like to avoid it. are
the
On Thu, Jan 19, 2012 at 10:48, Kenneth C. Arnold
wrote:
> On Thu, Jan 19, 2012 at 3:05 AM, Olivier Grisel
> wrote:
>> Rather than improving the error message when passing sparse arrays to
>> the dense impl of SVC we should refactor SVC to accept both dense and
>> sparse representation and use the
On Thu, Jan 19, 2012 at 3:05 AM, Olivier Grisel
wrote:
> Rather than improving the error message when passing sparse arrays to
> the dense impl of SVC we should refactor SVC to accept both dense and
> sparse representation and use the right wrapper as already done for
> SGD, LinearSVC, LogisticReg
2012/1/19 Andreas :
> I'll gladly review your pull request ;)
+1 :)
--
Lars Buitinck
Scientific programmer, ILPS
University of Amsterdam
--
Keep Your Developer Skills Current with LearnDevNow!
The most comprehensive onl
On 01/19/2012 09:05 AM, Olivier Grisel wrote:
> 2012/1/19 Gael Varoquaux:
>
>> On Thu, Jan 19, 2012 at 12:13:38PM +0900, Mathieu Blondel wrote:
>>
>>> Since your data is sparse, you need to use svm.sparse.SVC, not svm.SVC.
>>>
>> Those error messages are really not enlightning. Ma
2012/1/19 Stéfan van der Walt :
> Talking of which, I see in the current docs that coefficients can be
> specified to initialise, e.g., Lasso [1], but in the development
> version that is no longer possible. What is the new suggested way of
> doing warm starts?
Warm start is implemented in the d
On Thu, Jan 19, 2012 at 12:44 AM, Mathieu Blondel wrote:
> Here's a a pull-request implementing more convenient warm-start in SGD
> and ElasticNet:
>
> https://github.com/scikit-learn/scikit-learn/pull/568
Talking of which, I see in the current docs that coefficients can be
specified to initialis
Here's a a pull-request implementing more convenient warm-start in SGD
and ElasticNet:
https://github.com/scikit-learn/scikit-learn/pull/568
Comments welcome!
Mathieu
--
Keep Your Developer Skills Current with LearnDevN
To be pragmatic I would: go for MDP.
Gael
- Original message -
> >
> > I don't have the Bishop, and I must confess that I am still confused by
> > the Wikipedia. That said, it doesn't really matter. As long as people
> > feel confident that it is well defined and useful, it belongs to th
2012/1/19 Joris A. :
>
> In the meantime, any good libs to recommend? MDP?
Yes have a look at MDP and maybe also scikits.statsmodels which is
focused more classical statistics for finance and economics than
scikit-learn.
--
Olivier
http://twitter.com/ogrisel - http://github.com/ogrisel
>
> I don't have the Bishop, and I must confess that I am still confused by
> the Wikipedia. That said, it doesn't really matter. As long as people
> feel confident that it is well defined and useful, it belongs to the
> scikit, and I am all for it :).
>
> Gael
>
>
>
Thank you all!
Let's hope it w
2012/1/19 Alexandre Gramfort :
> +1 for FA. It's standard and indeed very similar to ProbabilisticPCA
> that we have.
>
> Now we need a volunteer :)
Is there a way to implement it in a scalable way (w.r.t n_samples and
n_features and n_factors / n_components)?
Because if it fallbacks to the defau
Hi list,
I'm more than +1 for online learning, it could be a killing feature of the
scikit !
I also like the first suggestion of Andreas, about Multinomial Logistic
regression. I think there is interesting work to do in the junction with
Bayesian statistics and priors.
Vincent
2012/1/19 Alex
+1 for FA. It's standard and indeed very similar to ProbabilisticPCA
that we have.
Now we need a volunteer :)
Alex
On Thu, Jan 19, 2012 at 7:31 AM, Gael Varoquaux
wrote:
> On Wed, Jan 18, 2012 at 11:53:01PM +0100, Andreas wrote:
>> > Factor analysis is a decomposition with a particular
>> > ass
2012/1/19 Gael Varoquaux :
> On Thu, Jan 19, 2012 at 12:13:38PM +0900, Mathieu Blondel wrote:
>> Since your data is sparse, you need to use svm.sparse.SVC, not svm.SVC.
>
> Those error messages are really not enlightning. Mathieu, you were saying
> in the thread about GSOC that sparse functionality
i've created the wiki page to organize what was suggested and so
people can volunteer for mentoring.
https://github.com/scikit-learn/scikit-learn/wiki/A-list-of-topics-for-a-google-summer-of-code-%28gsoc%29-2012
Alex
On Thu, Jan 19, 2012 at 8:38 AM, Peter Prettenhofer
wrote:
[..]
> - S
21 matches
Mail list logo
