home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!portal!cup.portal.com!lordSnooty
- From: lordSnooty@cup.portal.com (Andrew - Palfreyman)
- Newsgroups: comp.ai.neural-nets
- Subject: Re: Correctness of NNs
- Message-ID: <62758@cup.portal.com>
- Date: Fri, 24 Jul 92 19:49:57 PDT
- Organization: The Portal System (TM)
- References: <10088@baird.cs.strath.ac.uk>
- Lines: 23
-
- Robert B Lambert writes:
- [...]
- : Neural networks, like fuzzy logic have the advantage of being able to produce
- : sensible responses to new inputs. In handwritten OCR for example, reading a
- : character at a time be it by a human, ANN, or some other technique, must
- : give recognition rates of less than 100%. Everyone writes differently and
- : no system could be trained on examples of every human beings writing.
- [...]
-
- If you cast your mind back to the time when (oh, 1985/6 maybe) Expert Systems
- were the New Kids On The Block, you'll recall that NNs made a competitive
- splash by emphasising their "robustness" vis a vis ESs, especially in the
- marginal domains of the knowledge base, where ESs would suddenly and often
- catastrophically "lose it" while NNs would "gracefully degrade".
-
- These days it seems that all this is now taken for granted. The fuzzy logic/
- ES meld, however, seems to have reversed this somewhat.
-
- Apologies for the handwaving generalisations.
- --------------------------------------------------------------------------
- | lord snooty @the giant | how DARE they cancel Dennis Miller? |
- | poisoned electric head | andrew palfreyman |
- --------------------------------------------------------------------------
-