hi ariel,

have you standardized your data?

did you try with Ridge?

Alex

On Fri, May 25, 2012 at 7:26 PM, Ariel Rokem <[email protected]> wrote:
> Hi everyone,
>
> I am trying to use the SGDRegressor to solve for a sparse set of
> linear equations. I am getting an under/over-flow error (see below).
>
> Running master from github from this morning on Fedora 14.
>
> Any ideas?
>
> Thanks!
>
> Ariel
>
> In [30]: X
> Out[30]:
> <533100x18788 sparse matrix of type '<type 'numpy.float64'>'
>        with 52124700 stored elements in Compressed Sparse Column format>
>
> In [31]: y
> Out[31]: array([  4.1 ,  42.1 ,  14.1 , ...,  11.82, -11.18,  25.82])
>
> In [32]: y.shape
> Out[32]: (533100,)
>
> In [33]: from sklearn.linear_model import SGDRegressor as SGD
>
> In [34]: S1 = SGD()
>
> In [35]: S1.fit(X,y)
> ---------------------------------------------------------------------------
> ValueError                                Traceback (most recent call last)
> /white/u6/arokem/<ipython-input-35-1ebf5d8bf9f4> in <module>()
> ----> 1 S1.fit(X,y)
>
> /home/arokem/usr/local/lib64/python2.7/site-packages/scikit_learn-0.11_git-py2.7-linux-x86_64.egg/sklearn/linear_model/stochastic_gradient.pyc
> in fit(self, X, y, coef_init, intercept_init, sample_weight)
>    705
>    706         return self._partial_fit(X, y, self.n_iter, sample_weight,
> --> 707                                  coef_init, intercept_init)
>    708
>    709     def decision_function(self, X):
>
> /home/arokem/usr/local/lib64/python2.7/site-packages/scikit_learn-0.11_git-py2.7-linux-x86_64.egg/sklearn/linear_model/stochastic_gradient.pyc
> in _partial_fit(self, X, y, n_iter, sample_weight, coef_init,
> intercept_init)
>    640                                          coef_init, intercept_init)
>    641
> --> 642         self._fit_regressor(X, y, sample_weight, n_iter)
>    643
>    644         self.t_ += n_iter * n_samples
>
> /home/arokem/usr/local/lib64/python2.7/site-packages/scikit_learn-0.11_git-py2.7-linux-x86_64.egg/sklearn/linear_model/stochastic_gradient.pyc
> in _fit_regressor(self, X, y, sample_weight, n_iter)
>    754                                           self.learning_rate_code,
>    755                                           self.eta0,
> self.power_t, self.t_,
> --> 756                                           intercept_decay)
>    757
>    758         self.intercept_ = np.atleast_1d(intercept)
>
> /home/arokem/usr/local/lib64/python2.7/site-packages/scikit_learn-0.11_git-py2.7-linux-x86_64.egg/sklearn/linear_model/sgd_fast.so
> in sklearn.linear_model.sgd_fast.plain_sgd
> (sklearn/linear_model/sgd_fast.c:6164)()
>
> ValueError: floating-point under-/overflow occured.
>
> ------------------------------------------------------------------------------
> 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

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