hmm interesting it seems that using only low level svms it worked this
time.. but broke on LabelSpreading. Also please uncomment the higher degree
SVM's I think those are the ones that tend to crash. let me test again
also. ... and I forgot to use ones with the poly kernel ... I think that's
the problem
---------------
working on DecisionTreeClassifier
DecisionTreeClassifier difference: [ 0.01] percent diff: [ 0.11890606]
working on SVM:sigmoid:3
SVM:sigmoid:3 difference: [ 0.41] percent diff: [ 4.87514863]
working on SVM:rbf
SVM:rbf difference: [ 0.41] percent diff: [ 4.87514863]
working on SVM:sigmoid:4
SVM:sigmoid:4 difference: [ 0.41] percent diff: [ 4.87514863]
working on LASSO:0.1
LASSO:0.1 difference: 0.247515523301 percent diff: 2.94310967064
working on NaiveBayes
NaiveBayes difference: [ 0.11] percent diff: [ 1.30796671]
working on DecisionTreeRegressor
DecisionTreeRegressor difference: [ 0.01] percent diff: [ 0.11890606]
working on LabelSpreading:rbf
Traceback (most recent call last):
File
"C:\Users\Shomiron\Documents\GitHub\Sandbox\Misc\Allthemodels\sim_data.py",
line 85, in <module>
serial()
File
"C:\Users\Shomiron\Documents\GitHub\Sandbox\Misc\Allthemodels\sim_data.py",
line 81, in serial
e.fit(X,Y)
File
"C:\Python27\lib\site-packages\sklearn\semi_supervised\label_propagation.py",
line 211, in fit
graph_matrix = self._build_graph()
File
"C:\Python27\lib\site-packages\sklearn\semi_supervised\label_propagation.py",
line 383, in _build_graph
affinity_matrix = self._get_kernel(self.X_)
File
"C:\Python27\lib\site-packages\sklearn\semi_supervised\label_propagation.py",
line 115, in _get_kernel
return rbf_kernel(X, X, gamma=self.gamma)
File "C:\Python27\lib\site-packages\sklearn\metrics\pairwise.py", line
347, in rbf_kernel
K = euclidean_distances(X, Y, squared=True)
File "C:\Python27\lib\site-packages\sklearn\metrics\pairwise.py", line
174, in euclidean_distances
distances = safe_sparse_dot(X, Y.T, dense_output=True)
File "C:\Python27\lib\site-packages\sklearn\utils\extmath.py", line 78,
in safe_sparse_dot
return np.dot(a, b)
MemoryError
>>>
On 3 December 2012 00:36, Ronnie Ghose <[email protected]> wrote:
> oh hmm so if anyone wants to suggest additions to the script i'm all for
> it. it's a horrible mess :) and i didn't have any success in running the
> classifiers in parallel. ironically due to what i thought was an error
> caused by it (which i now find to be the svm). ... besides things like
> feature scaling ~ this is just a test one.
>
> also anyone know any way to speed it up :P ? cross validations assume
> you're running multiple in different threads? .... no?
>
> run it overnight, you should see a seg fault on one of the svms when you
> wake up
>
> version: 12.1
>
> https://github.com/RONNCC/Sandbox/tree/master/Misc/Allthemodels just run
> sim_data.py
>
> -- backupish ->
>
> Script: http://pastie.org/5465841
> data: ok so I don't know how to make this better so here's the whole
> thing, the subsets don't always generate the error and as i try more it of
> course grows gigantic ._.
>
>
> thanks
>
>
>
>
>
> On 2 December 2012 04:22, Mathieu Blondel <[email protected]> wrote:
>
>>
>> On Sun, Dec 2, 2012 at 5:44 PM, Ronnie Ghose <[email protected]>wrote:
>>
>>> So somehow while running an SVM model i'm getting a nice seg fault :(.
>>>
>>> http://pastie.org/5465679 is the last 50 lines of the strace file.....
>>>
>>> in specific it seems to be SVM with poly setting and degree=10. I'm
>>> running it in 20000 numbers... then I tried to lower it to 1000 and still
>>> got seg fault..
>>>
>>
>> To help us reproduce your problem, it would be nice to give us:
>> - the version of scikit-learn you're using
>> - the script you're using
>> - the data you're using (ideally a minimalistic subset that triggers the
>> bug)
>>
>> Thanks,
>> Mathieu
>>
>>
>> ------------------------------------------------------------------------------
>> Keep yourself connected to Go Parallel:
>> DESIGN Expert tips on starting your parallel project right.
>> http://goparallel.sourceforge.net/
>> _______________________________________________
>> Scikit-learn-general mailing list
>> [email protected]
>> https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
>>
>>
>
------------------------------------------------------------------------------
Keep yourself connected to Go Parallel:
BUILD Helping you discover the best ways to construct your parallel projects.
http://goparallel.sourceforge.net
_______________________________________________
Scikit-learn-general mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/scikit-learn-general
