svm - Is there a way to convert discriminant values in SVMLight multi-class classification into probability scores -
i using svm light multi-class classifier training classifier 4 classes. in classification stage classifier outputs predicted label , scores 4 classes. svm light website says, these scores "the discriminant values each of k classes". want show probability value of each of class users. wondering if there mathematical trick or other way using can "convert" these values probability values or @ least normalised score in between 0 , 1?
svm won't lead probability estimates. there tricks doing so, require additonal models, example platt's scaling.
on other hand, if want "scale" arbitrary number [0,1] interval can use sigmoid function so.
Comments
Post a Comment