home *** CD-ROM | disk | FTP | other *** search
- Xref: sparky comp.ai.neural-nets:4330 uw.neural-nets:3
- Newsgroups: comp.ai.neural-nets,uw.neural-nets
- Path: sparky!uunet!zaphod.mps.ohio-state.edu!darwin.sura.net!jvnc.net!nuscc!gold!tim
- From: tim@gold.iss.nus.sg (Tim Poston)
- Subject: Re: How deep are spurious attractor basins?
- Message-ID: <1992Nov20.064220.25157@nuscc.nus.sg>
- Keywords: Hopfield, Boltzmann, Attractors, Energy Functions
- Sender: usenet@nuscc.nus.sg
- Reply-To: tim@iss.nus.sg (Tim Poston)
- Organization: Institute of Systems Science, NUS, Singapore
- References: <Bxv79K.G7F@watserv1.uwaterloo.ca>
- Date: Fri, 20 Nov 1992 06:42:20 GMT
- Lines: 19
-
- In article <Bxv79K.G7F@watserv1.uwaterloo.ca> joordens@watserv1.uwaterloo.ca (Steve Joordens) writes:
-
- > However, I
- >have heard that this may not be the case (i.e., that spurious attractors
- >may be as deep as learned attractors).
-
- If "as deep as" means "a sum-of-squared errors that is just as low",
- it also surely means "as good as".
- In this case, what is "spurious" about such attractors?
-
- If you have a better function for measuring goodness,
- use it in learning.
-
- Tim
-
-
- ___________________________________________________________________
- My lucky number is irrational.
- ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
-