Edited by author 04-11-2002 04:18 AM
I agree with Dave's comment on the lack of explanation of domain specific vocabulary. I was particularly hung up on the paper's use of the term "VC dimension", which the authors do not define (and of which I was not familiar). A one sentence definition of the Vapnik-Chervonenkis dimension would have been quite helpful. I don't think the authors were very concerned with making this paper accessible to readers without a strong background in statistical learning theory.