Multiplicative update uses the vector w of an SVM to do feature selection.
At each iteration an svm is trained and the data is multiplied by the
weight vector of the classifier.
J. Weston, A. Elisseeff, M. Tipping and B. Scholkopf.
Use of the zero norm with linear models and kernel methods.
JMLR special Issue on Variable and Feature selection, 2002.
|
__init__(self,
arg=None,
**settings)
x.__init__(...) initializes x; see x.__class__.__doc__ for signature |
source code
|
|
|
|
|
|
|
|
|
|
|
selectFeatures(self,
data,
*options,
**args)
XXX for multi-class -- do one against the rest
and use the absolute value of the average/maximum value of w to rescale
multi-class |
source code
|
|
Inherited from object :
__delattr__ ,
__getattribute__ ,
__hash__ ,
__new__ ,
__reduce__ ,
__reduce_ex__ ,
__setattr__ ,
__str__
|
|
rank(self,
data,
**args)
Returns:
a ranking of the features in the dataset by converting the scores
to ranks |
source code
|
|
|
score(self,
data,
**args)
Returns:
a score for each feature in the input dataset |
source code
|
|
|
select(self,
data,
*options,
**args)
invokes selectFeatures to find predictive features and eliminates
the rest of the features from the input dataset |
source code
|
|
|
|
|
train(self,
data,
*options,
**args)
invokes selectFeatures to find predictive features and eliminates
the rest of the features from the input dataset |
source code
|
|