machine learning - SVM Scores - Upper Bound? -
I have concerns about supporting vector machines, that is, their classification score: Is there an upper limit on the score of these classifications? I do not think, because a SVM is only a hyperplane, and the score is basically the distance of a point from that hyperplane. Without restriction, a point can be anywhere, and thus there is no obligation in the distance, right?
I am asking because I have read the following line:
"When decision scores are surrounded - and SVM is surrounded by margin ... ... "
Can you explain it? I do not know how it is bound to know the fringe score ...
Thank you for your help, I appreciate it!
Your intuition is correct Whatever misrepresentation you misread is best or simply wrong (any case Some reference is necessary). SVM decision is not upper bound in values. It relies entirely on test instances.
The SVM function is a linear combination of internal products in the Value Spacing Value Support feature's feature space and support vectors. If the test examples are endless, then these internal products will also be unlimited.
Comments
Post a Comment