DM: RE: Data Forms for Mining (Limit on variables)

 > From: Frank Buckler <buckler@m2.uni-hannover.de>
 > ...
 > The vc-dimension tells how many example a specific learning algorithm can
 > learn/represent. In order to achieve ANY generalisation you need more
 > examples than the vc-dimension.

That is not correct. VC theory provides worst-case distribution free
bounds. It is often found in practice that the VC bounds are
extremely pessimistic. Current research is leading away from the VC
dimension to other concepts such as fat-shatttering dimension that
are more relevant to practical applications.

--

Warren S. Sarle       SAS Institute Inc.   The opinions expressed here
saswss@unx.sas.com    SAS Campus Drive     are mine and not necessarily
(919) 677-8000        Cary, NC 27513, USA  those of SAS Institute.

• Prev by Date: RE: DM: RE: Data Forms for Mining
• Next by Date: DM: AW: Classification problem
• Prev by thread: RE: DM: RE: Data Forms for Mining (Limit on variables)
• Next by thread: DM: Archive of the list
• Index(es):
  • Date
  • Thread