home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.ai.neural-nets
- Path: sparky!uunet!zaphod.mps.ohio-state.edu!saimiri.primate.wisc.edu!ames!purdue!yuma!longs.LANCE.ColoState.Edu!irakliot
- From: irakliot@longs.LANCE.ColoState.Edu (Leo J. Irakliotis)
- Subject: ART rotationally invariant ?
- Sender: news@yuma.ACNS.ColoState.EDU (News Account)
- Message-ID: <Dec31.082038.2447@yuma.ACNS.ColoState.EDU>
- Date: Thu, 31 Dec 1992 08:20:38 GMT
- Reply-To: irakliot@lance.colostate.edu
- Nntp-Posting-Host: engineer.lance.colostate.edu
- Organization: Engineering Network Services, Colorado State University
- Lines: 17
-
-
- While an ART(2) type NN can classify a vector X under the same class
- with the vector aX, a:real number, I was wondering if there is any
- work done on turning an ART network to be rotationally invariant,
- ie to classify vectors X and aX+T where T=[t t t ... t] under the
- same class.
-
- Any suggestion welcome.
-
- Best wishes for a happy (and productive) new year,
-
- --
- Leo J Irakliotis irakliot@eva.lance.colostate.edu
- -----------------------------------------------------------------
- Electrical Engineering l.irakliotis@ieee.org
- Colorado State U
- (303) 491-2021 Optical Computing Lab
-