home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!zaphod.mps.ohio-state.edu!wupost!uwm.edu!ogicse!das-news.harvard.edu!cantaloupe.srv.cs.cmu.edu!crabapple.srv.cs.cmu.edu!news
- From: sef@sef-pmax.slisp.cs.cmu.edu
- Newsgroups: comp.ai.neural-nets
- Subject: Re: re:need for unique test sets
- Message-ID: <1992Jul27.135653.270071@cs.cmu.edu>
- Date: 27 Jul 92 13:56:53 GMT
- Article-I.D.: cs.1992Jul27.135653.270071
- Organization: School of Computer Science, Carnegie Mellon
- Lines: 28
- Nntp-Posting-Host: sef-pmax.slisp.cs.cmu.edu
-
-
- From: tom@cs.su.oz.au (Thomas James Jones)
-
- I guess this is pretty difficult to answer, but what
- basis do we have for expecting that the net should
- (or can) learn a function we have 'in mind', when there
- are infinite trivially different ways of mapping any
- given data set?
- If we have no organizational basis for our learning, then
- what inherent organizational concepts do we expect the
- net to extract?
-
- Yes, it is pretty difficult to answer. Any given architecture or algorithm
- will be "biased" in favor of certain representations for a set of data, and
- this bias may be better or worse for generalization to new problems drawn
- from the same distribution. The state of theory in this area can generally
- be found in the COLT conference proceedings. That's "Conference on
- Learning Theory", and is not restricted to neural net models.
-
- -- Scott
- ===========================================================================
- Scott E. Fahlman
- School of Computer Science
- Carnegie Mellon University
- 5000 Forbes Avenue
- Pittsburgh, PA 15213
-
- Internet: sef+@cs.cmu.edu
-