Delivery-Date: 
Sender: info-hol-request@leopard.cs.byu.edu
Errors-To: info-hol-request@leopard.cs.byu.edu
Precedence: bulk
Date: Thu, 18 Nov 93 21:45:03 PST
From: "peter v. homeier" <homeier@cs.ucla.edu>
Message-Id: <931119.054503z.03483.homeier@oahu.cs.ucla.edu>
To: mel@ultrastar.EE.CORNELL.EDU (Miriam Leeser), info-hol@cs.uidaho.edu
Cc: homeier@cs.ucla.edu
Subject: Re: Is this a joke?
In-Reply-To: Message of Thu, 18 Nov 93 20:09:16 EST from "mel@ultrastar.EE.CORNELL.EDU (Miriam Leeser)" <9311190109.AA24066@ultrastar.EE.CORNELL.EDU>

It certainly is odd, but it is just barely possible that it may not be
a joke.  The Japanese are very proud of their involvement with fuzzy
logic.  I have done some small amount of research in this field, and
I believe I can understand what Mr. Bagubayashi may be referring to.

Basically, fuzzy logic is an extension of normal, classical logic in
that the truth value of a statement is not a member of {true,false},
but rather a value in the closed real interval from 0.0 to 1.0.  Here
1.0 represents complete truth, 0.0 represents complete falsity, and
intermediate values represent degress of truth, "shades of gray".

Invented by Lotfi Zadeh 25 years ago, this was adapted to manage
uncertainty in reasoning.  Rules were invented to calculate the results
of fuzzy inferencing, using both fuzzy rules and fuzzy statements to
match the rules.

In its early years, this field was widely derided as absurd and useless.
This attitude has persisted in the West.  However, principally in Japan,
in recent years some of these ideas have been applied to control problems,
such as controlling the accelleration or decelleration of a train.  The
results have been impressive, in that the control systems resulting are
more smooth and flexible than the traditional control systems, and attain
their goals quicker without overcorrection.  An example would be the 
Sendai subway, which decellerates much more accurately than either human
or traditional control systems could, matching marks on the pavement at
the station within a quarter inch, every time.  This is achieved with
such smooth decelleration that even standing passengers do not need to
hold on to supports.

These successes have lead some, including Bart Kosko, to reject Western
traditions of true/false dichotomies in favor of a continuous scale,
which seems more attuned to the Eastern tradition.  However, this
mixture of Zen philosophy with computer science makes the use of fuzzy
logic even more questionable to Western minds.

What is most fundamentally indetermined here is the actual meaning or
formal semantics of a truth value between 0.0 and 1.0.  In the papers
I have read, I have not seen any satisfactory definition of this semantics.
Given this, the formal semantics of the rest of the system is also not
established.

It is interesting to note that the systems that have been so successful
in fact do not use the fuzzy reasoning systems prevalent in the theoretical
community.  Rather, they use a simpler system which allows non-linear
control response to input variables.  The reasoning is very simple, just
one step from input sensor variables to output actuator values.  The
non-linearity gives the control needed for the better response, with the
fuzzy calculations smoothly melding the different parts of the control
surface together.

> "We plan to be the first to apply the technology to aeronautics, where
> it is perfect for situations such as determining about where the flaps
> are positioned, whether or not the plane is more or less pointing at the
> runway during landing, and  whether or not the plane is sort of going in
> the right direction."

The final determination of the worth of fuzzy logic is yet to be seen.
They may able to apply these techniques to aircraft control, but it is
also possible that the Japanese may be overreaching themselves here.

If anyone is interested, I can send them a postscript version of a paper
I presented this year on fuzzy inferencing.

Peter Homeier
homeier@cs.ucla.edu