From dove!uunet!wupost!sdd.hp.com!sgiblab!nec-gw!nec-tyo!wnoc-tyo-news!ccut!wsclark!lab!lab!farzin Wed Oct 14 12:50:48 MDT 1992
Article: 4658 of comp.ai
Path: dove!uunet!wupost!sdd.hp.com!sgiblab!nec-gw!nec-tyo!wnoc-tyo-news!ccut!wsclark!lab!lab!farzin
From: farzin@apollo3d.apollo3.ntt.JP (Farzin Mokhtarian)
Newsgroups: comp.ai,sci.engr.control
Subject: Fuzzy Logic in Japan
Message-ID: <1so.hkice@lab.ntt.jp>
Date: 7 Oct 92 23:14:54 GMT
References: <524@ssdc.UUCP> <1992Oct4.234858.16404@samba.oit.unc.edu>
	<84475@bcsaic.boeing.com>
Sender: news@lab.ntt.jp
Organization: none
Lines: 139
Xref: dove comp.ai:4658 sci.engr.control:157



  
This article is about Fuzzy Logic and appeared a while ago in
the Japan Times. Although it is not at a very technical level,
it might provide some insight on the state of Fuzzy Logic in Japan.
-------------------------------------------------------------------
  
From: Japan Times, July 20, 1992
   
Fuzzy logic best summed up by common sense
  
Computer Corner 
John Boyd
   
Fuzzy logic was introduced to the world 27 years ago by Professor
Lotfi Zadeh in his "Fuzzy Sets" paper published in Information
Control magazine, though it is only recently that we've seen it
applied across a broad range of products.
    
Some readers have asked for more explanation on fuzzy logic, so
here's an attempt to "defuzzify" the subject a little further.
   
Simply put, fuzzy logic is aimed at enhancing our prissy computer
technology with a touch of common sense.
    
One problem with the conventional digital computer is that it is
such a scrupulously either-or beast. It cannot be easily coaxed
to handle approximations or vague notions like young, a lot and
probably.
   
Yet most of us rely on such terms daily because we happen to be
humans dealing with other humans, not robots building cars.
    
It's an easy matter to arbitrarily program a computer so it
designates everyone falling into the age-range 0f 15 to 18
as being a youth. Such a precise category has come to be called
a crisp set since the emergence of fuzzy logic.
    
Yet we all know some 14-year-olds can look older than some
late-developers turning 20. Such exceptions, however, cannot
be accounted for in conventional computing. Or at least not
without an inordinate amount of additional programming and
expense.
    
As Tetsuya Yamada, a senior engineer at Hitachi Ltd., replied
when I asked him if we couldn't just continue using conventional
programming and technology for controlling new products, instead
of fuzzy, "Well, we could. And you could probably swim across the
Pacific if you got enough support from enough people. But ..."
   
To overcome this problem, Zadeh was inspired to develop his fuzzy
theory and the math to go with it that could be used to create
fuzzy sets based on imprecise natural language.
   
Each member in a fuzzy set (such as the youths and others considered
in the above example) is assigned one of a continuous range of values
(called the membership value) between zero and one.
   
Whereas in the above crisp set a 13-year-old going on 14 would still
have to be considered a minor and thus be designated as zero in
binary logic, fuzzy logic could assign him a membership value of
say 0.1. Likewise, an immature 20-year-old who would normally fall
outside our either-or crisp-set range could be assigned a membership
value of 0.9 depending upon the criteria we use to measure youth.
   
Working out just what criteria to use, what values should be assigned
each member and deciding what rules are necessary to govern the
relationships between members is the key to successfully applying
fuzzy control in products. 

In some applications, determining the optimum rules has become so
complex, some manufacturers have resorted to employing the aid of
neural networks, which may be stretching a good thing too far, given
fuzzy logic's original purpose to get round complexity.
   
Still, the flexibility in herent in fuzzy is clearly useful in
dealing with approximate calculations, such as "about 100".
    
It can be used in artificial intelligence to provide us with an
"almost true" answer. It can also infer a common-sense result even
when the data is not precise.
       
Our handwritten 5 in 250 would be treated as "5," not the letter "S,"
for instance, in Sony's fuzzy-based Palmtop computer.
   
While we have all seen fuzzy logic-based products from the likes of
Matsushita, Sanyo and Hitachi, one unlikely company that has made
fuzzy technology a central part of its business strategy is Omron
Corp.
      
It began its research into fuzzy logic in 1984 and has since applied
for over 700 patents. This puts it in the forefront of fuzzy
applications in areas like factory and industry control, as well as
in medical equipment.
   
In 1989, Omron also signed on lotfi Zadeh as a senior advisor.
   
Earlier this year at the Business Show in Harumi, Omron demonstrated
its fuzzy workstation. Omron manufactures both standard Motorola
68040-based and 88000 reduced-instruction or RISC-based workstations
that can be fitted with a fuzzy inference board, turning them into
the world's first fuzzy workstations.
      
Omron claims such a RISC-based workstation can achieve 4 billion
operations per second, an incredible speed if they haven't fuzzed
on the number. Fuzzy logic is used in the workstations to store
and retrieve fuzzy information and make inferences. 
   
Ranging in price from Y2.5 million to almost Y4 million (a US dollar
is about 120 Yens  -FM), these machines are not the kind of products
you will find down in Akihabara. (a section of Tokyo famous for its
quantity and variety of electronic goods  -FM) Rather, they are
typically aimed at value-added resellers in niche markets, and
engineers who want to develop fuzzy applications, fuzzy databases
and expert systems, as well as fuzzy inference systems.
     
However, the entrepreneurs among you may be interested in Omron's
FB-30AT fuzzy inference board for the IBM PC and compatible wares.
It features a 24 MHz FP-3000 fuzzy chip capable of processing up
to 128 rules, with five antecedents and 2 consequents. Training
software and a compiler is also available.
      
Omron has also produced a fine little booklet on fuzzy called
"Clearly Fuzzy" that I dipped into when writing this column.
      
Tadashi Katsuno, at Omron's public relations section, tells me
he still has a limited number of copies left that he will send
to the first readers of Computer Corner who write to him with
contact information.
   
The address is Omron Corp., International Public Relations Section,
Omron Tokyo Bld., 3-4-10 Toranomon, Minato Ward, Tokyo 105.
--------------------------------------------------------------------
   
 - Farzin Mokhtarian
   farzin@apollo3.ntt.jp
                    
                   


