Hi Thomas,
the doc says, that nu gives an upper bound on the fraction of training errors
and a lower bound of the fractions
of support vectors.
http://scikit-learn.org/stable/modules/generated/sklearn.svm.NuSVC.html
Therefore, it acts as a hard bound on the allowed misclassification on your
dataset.
To me it seems as if the error bound is not feasible.
How well did the SVC perform? What was your training error there?
Will the NuSVC converge when you skip the sample_weights?
Greets,
Piotr
On 08.12.2016 00:07, Thomas Evangelidis wrote:
Greetings,
I want to use the Nu-Support Vector Classifier with the following input data:
X= [
array([ 3.90387012, 1.60732281, -0.33315799, 4.02770896,
1.82337731, -0.74007214, 6.75989219, 3.68538903,
..................
0. , 11.64276776, 0. , 0. ]),
array([ 3.36856769e+00, 1.48705816e+00, 4.28566992e-01,
3.35622071e+00, 1.64046508e+00, 5.66879661e-01,
.....................
4.25335335e+00, 1.96508829e+00, 8.63453394e-06]),
array([ 3.74986249e+00, 1.69060713e+00, -5.09921270e-01,
3.76320781e+00, 1.67664455e+00, -6.21126735e-01,
..........................
4.16700259e+00, 1.88688784e+00, 7.34729942e-06]),
.......
]
and
Y= [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
............................
0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0]
Each array of X contains 60 numbers and the dataset consists of 48 positive
and 1230 negative observations. When I train an svm.SVC() classifier I get
quite good predictions, but wit the svm.NuSVC() I keep getting the following
error no matter which value of nu in [0.1, ..., 0.9, 0.99, 0.999, 0.9999] I try:
/usr/local/lib/python2.7/dist-packages/sklearn/svm/base.pyc in fit(self, X, y,
sample_weight)
187
188 seed = rnd.randint(np.iinfo('i').max)
--> 189 fit(X, y, sample_weight, solver_type, kernel, random_seed=seed)
190 # see comment on the other call to np.iinfo in this file
191
/usr/local/lib/python2.7/dist-packages/sklearn/svm/base.pyc in _dense_fit(self,
X, y, sample_weight, solver_type, kernel, random_seed)
254 cache_size=self.cache_size, coef0=self.coef0,
255 gamma=self._gamma, epsilon=self.epsilon,
--> 256 max_iter=self.max_iter, random_seed=random_seed)
257
258 self._warn_from_fit_status()
/usr/local/lib/python2.7/dist-packages/sklearn/svm/libsvm.so in
sklearn.svm.libsvm.fit (sklearn/svm/libsvm.c:2501)()
ValueError: specified nu is infeasible
Does anyone know what might be wrong? Could it be the input data?
thanks in advance for any advice
Thomas
--
======================================================================
Thomas Evangelidis
Research Specialist
CEITEC - Central European Institute of Technology
Masaryk University
Kamenice 5/A35/1S081,
62500 Brno, Czech Republic
email: [email protected]<mailto:[email protected]>
[email protected]<mailto:[email protected]>
website: https://sites.google.com/site/thomasevangelidishomepage/
_______________________________________________
scikit-learn mailing list
[email protected]<mailto:[email protected]>
https://mail.python.org/mailman/listinfo/scikit-learn
_______________________________________________
scikit-learn mailing list
[email protected]
https://mail.python.org/mailman/listinfo/scikit-learn
