recognition rate generalization识别率 泛化

 

http://www1.inf.tu-dresden.de/~ds24/lehre/ml_ws_2013/ml_11_hinge.pdf

 

 

Two extremes:

 

• Big 𝐶 → the loss is more important → better recognition rate but smaller margin (worse generalization)

 

• Small 𝐶 → the generalization is more important → larger margin (more robust classifier) but worse recognition rate