Transformation Techniques on SVM

[avagreyyy4]

Hi! I understand we have learned that centering or scaling the data has no effect on the VC dimension since it is not impacting the dimensionality. However, for the SVM where we now have a dvc that relies on the radius and rho, would centering the data, for example, impact the dvc since it is decreasing the radius value? Thanks

[mikeizbicki]

You're correct that centering will reduce the bound on the "vc dimension" for SVM provided by Theorem 8.5 in the textbook. But you are also responsible for understanding why there are scare quotes around the phrase VC dimension above.