Ο Marcello Savarese <email@example.com> έγραψε στο μήνυμα συζήτησης:
> I read with attention you response about this subject....
> I'm agree with you considerations on argument, but I think that the
> imprecision is the natural aspect of fuzzy logic .
> I think that the uncertainty is linked whit probability concept.
> If I say that an object ( eg. apple) is 0.75( or 75%) an apple I expose a
> uncertainty but if I say this obj belongs 0.8 to apple membership, I
> means an imprecision.
> So in the probability topic I say that this obj is or isn't an apple
> grade of probability( in ex: 75%)) ( uncertainty), in the Fuzzy topic I
> that this obj is a "wrong apple" but is an apple( imprecision).
> I hope that I explained my concept in the better manner.....
> I love to say that probability is a distance measure from true and the
> is a distance measure from real.
> Where my reasoning is wrong?
> What do you think about it?
The following reply might be useful:
> Hi everyone,
> A neuron clearly showing the fuzzy logic INCLUSIVE OR property of passing
> maximum value among it inputs has been found in
> the auditory system of the frog brain (see the home page of my web site
> http://neurocomputing.org for the data and reference). Yet neural signals
> more than one parameter (dimension) unlike the values used in conventional
> and probability. Neural signals have a pulse width (number of action
> and within that pulse the action potentials have a frequency. In addition
> frequency has a variance and often a decay. Finally, a variable latency
> occurs before the pulse is triggered.
All true. Even more, R.Penrose suggests a rather quantum neural model for
the brain cells which takes into account quantum effects in signal
propagation, pointing to a theory as complex as (maybe) hyperstrings for
physics. However, this does not mean that current implementations a
deprecated (Newton's laws still stand, even as a sub-set, but quite real and
descriptive). With neural nets, we do not want to construct an exact
formulation for the biological neurons, but simply to simulate their basic
fuctionality. Neural networks seem to have accomplished that in many cases,
whether low-level signal processing or high-level logic functions. The
problem is not whether the NN can learn to act as a logic unit, but rather
how AI can help that this unit has the congitive knowledge of what it does.
> Certainly neural signals have a validity value in addition to a truth
> make it at least a 2-dimensional multivalued logic (a generalization of
> modal logic as pointed out by Stephen Lehmke in a posting to this
> how are these 2 dimensions used and defined? How do they interact? Is
> relation to probability? And what about their variances?
The vector space of a NN can actually contain many thousands of dimensions,
this does not mean that we should try and translate them all into
multi-dimensional logic. NN try to fit an adaptive system into a set of
pre-determined constraints (not necessarily logic ones), not describe the
cognitive attributes of the environment which is implied by them (NN is not
AI in terms of inference logic).
> Probability and logic and not independent. Consider this question about a
> belonging to the sample set of vehicles. "What is the probability that a
> behind a wall?" The probability depends upon the precision of the
> "car". Does "car" include a pick-up truck, a sport utility vehicle, a van?
I can not agree with that. Probability has nothing to do with fuzziness.
Propability applies to mutually exclusive states, while fuzziness applies
when things belong to more than one state at the same time. The proposition
"What is the probability that a car is behind a wall?" is clearly predicate
logic (with the addition of probability calculations). The same question in
terms of fuzziness should be "How much of the car is behind the wall?" - it
may sound strange but it stands quite well (with value from 0 to 1), as long
as we forget our perception that a car can not actually be in front and
behind the wall (two-valued logic => mutually exclusive states =>
Informatics Systems Analyst (MSc)
This message was posted through the fuzzy mailing list.
(1) To subscribe to this mailing list, send a message body of
"SUB FUZZY-MAIL myFirstName mySurname" to firstname.lastname@example.org
(2) To unsubscribe from this mailing list, send a message body of
"UNSUB FUZZY-MAIL" or "UNSUB FUZZY-MAIL email@example.com"
(3) To reach the human who maintains the list, send mail to
(4) WWW access and other information on Fuzzy Sets and Logic see
(5) WWW archive: http://www.dbai.tuwien.ac.at/marchives/fuzzy-mail/index.html
This archive was generated by hypermail 2b30 : Thu May 17 2001 - 15:33:36 MET DST