home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.ai.neural-nets
- Path: sparky!uunet!gatech!destroyer!ubc-cs!unixg.ubc.ca!kakwa.ucs.ualberta.ca!alberta!arms
- From: arms@cs.UAlberta.CA (Bill Armstrong)
- Subject: Re: need for unique test sets
- Message-ID: <arms.712338868@spedden>
- Sender: news@cs.UAlberta.CA (News Administrator)
- Nntp-Posting-Host: spedden.cs.ualberta.ca
- Organization: University of Alberta, Edmonton, Canada
- References: <25633@life.ai.mit.edu> <1992Jul22.031319.15531@afterlife.ncsc.mil> <1992Jul27.101401.18276@cs.su.oz.au> <1992Jul28.060228.7607@afterlife.ncsc.mil>
- Date: Tue, 28 Jul 1992 15:54:28 GMT
- Lines: 46
-
- hcbarth@afterlife.ncsc.mil (Bart Bartholomew) writes:
-
- > At the moment, we don't have any tools that I know
- >of to examine a BP net and understand what's going on except
- >for some trivial cases. I suspect I'd recognize an XOR net
- >fairly easily, but there are only four distinct configurations
- >for an XOR BP net. More complicated functions are more difficult
- >to analyze. Not necessarily impossible, but definately more
- >difficult. At this point I expect Prof Armstrong to point
- >to the inordinate ease with which an ALN can be analyzed.
-
- Why should I, especially after pointing out that the analysis problem is
- provably intractable (NP-complete). I just agreed with Scott Fahlman on this
- on the net yesterday.
-
- If you want to criticize me for something I actually advocate, why
- don't you take issue with my ideas for a "safe design methodology"?
-
- >I remain unconvinced, however, of the universal applicability
- >and/or superiority of ALNs over BP in all situations.
-
- I would expect you to remain unconvinced, at least until ALN software
- and hardware that is actually superior is available. I do try to
- reason about the extreme inefficiency of BP and how ALN superiority can
- be achieved though. People who want to follow the reasoning will do
- so; but most won't. Most of this will remain just that -- reasoning
- -- until adequate funding is available for the ALN approach, funding
- which is now being plowed into supporting arithmetic BP networks that
- have no future at all.
-
-
- One more question please, that maybe you are in a position to answer
- as a result of your (.mil) domain. I have done contracts for the
- Defence Research Establishment Valcartier, which showed that ALN ideas
- could be applied to FLIR images (ca. 1980). I heard from someone
- attached to a foreign military research establishment last year that
- my stuff was actually quite popular in their country. That surprised
- me a lot, since it didn't seem like the ALN idea had gotten very far.
- Do you know of any other non-classified work on ALNs that has been
- done for the military, that you could point me to?
-
- --
- ***************************************************
- Prof. William W. Armstrong, Computing Science Dept.
- University of Alberta; Edmonton, Alberta, Canada T6G 2H1
- arms@cs.ualberta.ca Tel(403)492 2374 FAX 492 1071
-