[
Lists Home |
Date Index |
Thread Index
]
Hi Thomas,
Thomas said:
That is not new, BTW. Bart Kosko discussed a fuzzy resolution to the
Liar's Paradox on page 4 of his 1992 book "Neural Networks and Fuzzy
Systems".
Didier replies:
Yes but it project a new light on that type of problem. As you know, the
actual focus of W3 is on publishing not on building agent (at least from the
type of proposed solutions). When you want to build an agent tat will
process the proposition published then you need to find ways to resolve
conflicts.
But even with possibility theory (based on fuzzy sets) the problem of
assigning a value still remains. Some Machine learning researchers are using
game theory to provide such values. I do not know yet if an agent should use
a mix of game theory and possibility theory to at least produce something
useful our of propositions published on the web. I discovered that Google is
having some interests In conceptual clustering, so it seems that them too
are moving toward some fuzzy logic resolution. Anyway, I am working hard to
include weight value (ownership function values) to PDML so that it could be
used no only to transport Object or prototypes but also proposition about
concepts (a concept being equivalent to a class).
Cheers
Didier PH Martin
http://Didier-martin.com
|
