DOC: Removed an unsubstantited claim about AVDD generalising the ocSVM [ci skip]

ivannz · ivannz · commit 5f5dc2939a87 · 2016-11-21T01:48:29.000+03:00
diff --git a/doc/modules/svm.rst b/doc/modules/svm.rst
@@ -779,8 +779,7 @@ typical run of the SVDD model the weights are set to :math:`w_i = 1`, which
 is equivalent to the original 2013 SVDD problem for :math:`C = \frac{1}{\nu l}`.
 
 In the case of a stationary kernel :math:`K(x,y)=K(x-y)` the SVDD model
-and the One-Class SVM models are provably equivalent, which means that SVDD
-generalizes the former.
+and the One-Class SVM models are provably equivalent.
 
 The primal problem of Scikit's version of SVDD for the training sample
 :math:`(x_i)_{i=1}^l` with weights :math:`(w_i)_{i=1}^l` is:
