On Fri, May 11, 2012 at 10:47 AM, Gael Varoquaux <
gael.varoqu...@normalesup.org> wrote:
>
>
> $ nosetests --exe sklearn
>
ok... thanks so much. it now does not error out, though it has some of the
same AttributeError stuff:
Exception AttributeError: AttributeError("'NoneType' object has no
On Fri, May 11, 2012 at 10:44:45AM -0700, W. Bryan Smith wrote:
> Thank you Gaël, I can now install and import.
> The tests are failing though, log attached.
How did you run the tests? The recommended way is now to use
$ nosetests --exe sklearn
Using 'import sklearn; sklearn.test()' breaks fo
Thank you Gaël, I can now install and import.
The tests are failing though, log attached.
thanks,
bryan
On Fri, May 11, 2012 at 10:22 AM, Gael Varoquaux <
gael.varoqu...@normalesup.org> wrote:
> I could reproduce using an install rather than a build inplace, and I
> pushed a fix.
>
> Thanks for
I could reproduce using an install rather than a build inplace, and I
pushed a fix.
Thanks for reporting and sorry for the bug,
Gaël
On Fri, May 11, 2012 at 10:13:51AM -0700, W. Bryan Smith wrote:
> On Fri, May 11, 2012 at 3:03 PM, W. Bryan Smith wrote:
> >> the source was pulled from git abou
apologies for the formatting... i've configured my account to receive the
digest and i don't know how to get back to the original message. response
in-line below:
Message: 6
Date: Fri, 11 May 2012 15:26:30 +0900
From: Mathieu Blondel
Subject: Re: [Scikit-learn-general] __check_build error on skl
2012/5/11 Mathieu Blondel :
>
>
> On Fri, May 11, 2012 at 11:08 PM, Lars Buitinck wrote:
>>
>>
>> Shouldn't you set the intercept_ as well?
>>
> Indeed. And there was a typo. Obviously it should be
>
> clf.coef_ += clf2.coef_
But as demonstrated in my previous message, this won't work as
clf.coef
On Fri, May 11, 2012 at 11:08 PM, Lars Buitinck wrote:
>
> Shouldn't you set the intercept_ as well?
>
> Indeed. And there was a typo. Obviously it should be
clf.coef_ += clf2.coef_
Mathieu
--
Live Security Virtual Conf
On Fri, May 11, 2012 at 11:15 PM, Olivier Grisel
wrote:
>
> +1 : the semantics of warm_start is *only to speedup the convergence*
> by starting from a solution closer to the optimal solution of the
> convex optimization problem (in this case the final solution will be
> the solution of fit(X_subse
2012/5/11 Gael Varoquaux :
> On Fri, May 11, 2012 at 10:47:19PM +0900, Mathieu Blondel wrote:
>> All algorithms which supports a warm_start constructor option should also
>> be usable similarly to partial_fit. For example:
>
>> from sklearn.linear_model import Lasso
>
>> clf = Lasso(war
On Fri, May 11, 2012 at 10:47:19PM +0900, Mathieu Blondel wrote:
>All algorithms which supports a warm_start constructor option should also
>be usable similarly to partial_fit. For example:
>from sklearn.linear_model import Lasso
>clf = Lasso(warm_start=True)
>clf.fit(X_subset
2012/5/11 Mathieu Blondel :
> Another idea is to learn a different classifier on each subset and use a
> mixture of the classifiers. As a mixture weight, a simple choice is 1 /
> n_mixtures.
>
> clf = LinearSVC()
> clf.fit(X_subset1, y_subset1)
> clf2 = LinearSVC()
> clf2.fit(X_subset2, y_subset2)
2012/5/11 Mathieu Blondel :
>
>
> On Fri, May 11, 2012 at 10:52 PM, Olivier Grisel
> wrote:
>>
>> Unfortunately I don't think you can assign coef_ on liblinear wrapper
>> models due to internal memory layout constraints.
>>
>
> Sure you can :)
>
> https://github.com/scikit-learn/scikit-learn/blob/
On Fri, May 11, 2012 at 10:52 PM, Olivier Grisel
wrote:
> Unfortunately I don't think you can assign coef_ on liblinear wrapper
> models due to internal memory layout constraints.
>
>
Sure you can :)
https://github.com/scikit-learn/scikit-learn/blob/master/sklearn/svm/base.py#L805
Mathieu
--
2012/5/11 Mathieu Blondel :
> All algorithms which supports a warm_start constructor option should also be
> usable similarly to partial_fit. For example:
>
> from sklearn.linear_model import Lasso
>
> clf = Lasso(warm_start=True)
> clf.fit(X_subset1, y_subset1)
> clf.fit(X_subset2, y_subset2)
> ..
All algorithms which supports a warm_start constructor option should also
be usable similarly to partial_fit. For example:
from sklearn.linear_model import Lasso
clf = Lasso(warm_start=True)
clf.fit(X_subset1, y_subset1)
clf.fit(X_subset2, y_subset2)
...
Another idea is to learn a different clas
On Fri, May 11, 2012 at 02:53:26PM +0200, Peter Prettenhofer wrote:
> Incremental learning is supported via ``partial_fit``, however, for
> supervised learning only ``SGDClassifier`` [1] supports it (it should
> be easy to add it to ``MultinomialNB`` too [2]).
> For clustering you should have a loo
Hi Rafael,
Incremental learning is supported via ``partial_fit``, however, for
supervised learning only ``SGDClassifier`` [1] supports it (it should
be easy to add it to ``MultinomialNB`` too [2]).
For clustering you should have a look at ``MinibatchKMeans`` [3] it
supports ``partial_fit`` too - a
2012/5/11 Rafael Calsaverini :
> Any of the algorithms implemented in scikit-learn can be incrementally
> trained?
All estimators that have a partial_fit method can be trained
incrementally. That excludes the text vectorizer, unfortunately, but
it includes SGDClassifier (approximate linear SVM/log
2012/5/11 Rafael Calsaverini :
> Any of the algorithms implemented in scikit-learn can be incrementally
> trained?
>
> Three particular things are interesting to me: classifying texts,
> unsupervised clustering analysis of texts and hierarchical clustering
> analysis of texts. But my set of texts i
Any of the algorithms implemented in scikit-learn can be incrementally
trained?
Three particular things are interesting to me: classifying texts,
unsupervised clustering analysis of texts and hierarchical clustering
analysis of texts. But my set of texts is just too big to load in memory
all at on
> The current idea would be to use n_clusters for all clustering
> algorithms and n_components
> for GMM.
+1
B
--
Live Security Virtual Conference
Exclusive live event will cover all the ways today's security and
threat
On Fri, May 11, 2012 at 11:49:50AM +0200, Andreas Mueller wrote:
> The current idea would be to use n_clusters for all clustering
> algorithms and n_components
> for GMM.
> Comments?
I agree with this choice. It needs to good through a phase of
deprecation, but I find that it is a good policy.
Hi everybody.
I recently opened an issue on renaming the clustering parameters:
https://github.com/scikit-learn/scikit-learn/issues/844
At the moment, the parameter in KMeans and MiniBatchKMeans and
SpectralClustering is called k,
and n_clusters in ward.
The number of cluster centers in GMM is ca
2012/5/10 JAGANADH G :
> Hi all
>
> Is there any way to get the TF-IDF value mapped with the word vector in
> sklearn.
>
> I would like to get output like
>
> w1 -> TF-IDF
> w2 -> TF-IDF
TF is sample-dependent but the IDF weights for each feature index are
stored as an array attribute named `idf_`
2012/5/11 Vlad Niculae :
> 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
>
Hi list,
I've started a work-in-progress PR on multinomial logistic regression
for the SGD module. You can find it here [1].
I would really appreciate your input - especially on issues such as
API, learning rate schedule, implementation.
thanks,
Peter
[1] https://github.com/scikit-learn/scikit-
I'm +1 for scikit-learn/scikit-learn-speed or
scikit-learn/scikit-learn-vbench (if that's what we intend to use for
performance regression tests).
BTW: please keep me posted w.r.t. your efforts - I would really like
to add some performance regression tests for the SGD module - I've the
feeling tha
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
scikit-
28 matches
Mail list logo
