home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!hela.iti.org!usc!merelo
- From: merelo@rana.usc.edu (Juan Julian Merelo)
- Newsgroups: comp.ai.neural-nets
- Subject: Re: Kohonen Algorithm used in ANSIM Neural Net Software
- Date: 6 Nov 1992 12:15:23 -0800
- Organization: University of Southern California, Los Angeles, CA
- Lines: 10
- Distribution: world
- Message-ID: <1dejorINNhj8@rana.usc.edu>
- References: <1992Nov6.123143.94@polar.etsiig.uniovi.es>
- NNTP-Posting-Host: rana.usc.edu
-
-
- That was the algorithm used in the first (ie, Biol. Cyb. 1983, don't remenber the
- exact quote) Kohonen papers. After that, it changed the distance to euclidean
- instead of angle, and skipped normalization. I think that the only difference is
- that you miss one dimension, and it may be slower, as you have to do one
- additional summation.
-
- See ya
-
- JJ
-