1Re: Fuzzy Practical Uses

From: P. Sarma (psarma@seas.upenn.edu)
Date: Wed Aug 15 2001 - 13:08:17 MET DST

  • Next message: jok: "fuzzy logic and automatic transmission"

    Dear Earl,

    It is not my direct field, the business algorithms of FLS ... yet the
    letting of the query to generate its own fuzzy linguistic variables (FLV) was
    extremely interesting. How is this done? For example demographic
    distributions: so if I input "Age" as a query (or field), does it
    formulate some superset of FLV's whic descend into subsethood on adding
    more information, such as "college undergrads", or so? This self-updating,
    clearly efficient, is rarely seen in the engineering systems FLC works.
    Purely data-based learning/training can be very computationally heavy.
    Could you elaborate, a little, on the base technique? I have some
    idea of your views on FLC per se from the earlier article on Adaptive FLC in
    IEEE Spectrum, which I found interesting; FLC is my thrust field in Soft
    Computing.

    The key one is your comments on "robustness" ... this
    is an issue in all forms of systems, from AI-based to crisp to stochastic
    to numericals. It is always one of the top 2-3 desirable things in any system.
    I am aware, that you have substantial thousands of hours of experience working
    on FLS, from the slow mainframe IBM/VAX/NCR/CDC days (which forced very
    efficient coding methods!), to our modern super-PC's and workstation networks.

    My point:

    We, who gravitate more towards dynamics and control, hence FLC \subset
    FLS, have enormous robustness issues, too. Over the course of the last 3-4
    years, mostly during my thesis work (applied FLC developed with GA and
    some original model-reducing-booster structures), I have done many
    extensive numerical experiments that support and strengthen the Robustness
    properties. For example, in one classical nonlinear reactor, the
    optimally-tuned FLC shows superb "graceful degradation" (the
    key-parameter perturbation dK stepwise increased for a full closed-loop
    tracking run), and eventually fails at a "perturbation" level
    (w.r.t. a key nominal model parameter) that is almost an order of magnitude
    higher than well-tuned linear controllers ... and in fact, clearly exceeds
    the robustness levels of one of the best crisp nonlinear methods (feedback
    linearisation), clearly.

    Do you have any quantitative, algebraic, fuzzy set-theoretic ways of
    defining these robustness levels - put another way, for sufficiently small
    systems, is there some [perhaps approximate] way of ~predicting the
    robustness? There is always this contentious issue with so-called crisp
    NLC personnel, about this claimed robustness ... and it can be seen (in
    FLC eg.) from the starting block, of the Mamdani's 1974 boiler FLC, and
    ever since.

    Any thoughts you have on this issue would be welcome.

    Pramit

    On Tue, 14 Aug 2001, Earl Cox wrote:

    > Well, of course, there is the entire realm of FuzzySQL, the ability to use
    > fuzzy sets in database queries, allowing queries like,
    >
    > select *
    > from projects
    > where duration is Long
    > and budget is very high
    > and number_of_revisions is small
    > and delta(planned_end,actual_end) is tiny;
    >
    > such fuzzy queries are much more robust and fault tolerant than traditional
    > Boolean queries. They can also return a compatibility index indicating how
    > closely the returned records match the intent of the query's semantics.
    > Sonalysts markets a Fuzzy Query product for databases and Excel spread
    > sheets. The code for a basic FuzzySQL processor can be found in my book
    > Fuzzy Logic for Business and Industry.
    >
    > And, of course you can use fuzzy logic as part of database domain (or
    > integrity) checks as well as triggers. Although I have read papers about
    > storing fuzzy grades of memberships in a database, in the real world this is
    > almost never practical due to the volatility of data. It is much better to
    > let the query generate the memberships. Also, what is young to marketing
    > might not be young to credit approval or human resources -- thus storing the
    > fact that someone's "young" in a database removes flexibility and robustness
    > and the ability to use the same data for multiple purposes.
    >
    > Why would you think that fuzzy logic would use any more computing resources
    > than any other kind of computational intelligence system (heuristic
    > searches, neural networks, genetic algorithms, etc.) There are many
    > real-time fuzzy systems in everyday use involving hundreds of rules and
    > dozens of fuzzy sets. I have built on-line fuzzy eCommerce cross-marketing
    > and CRM analysis systems running as Java applets that analyze, partition,
    > rank, and segment hundreds of incoming transactions per second in a web
    > server or an application server (such as BEA WLS).
    >
    > Yes. And it has been since the early 1980's.
    >
    > I suggest that, instead of throwing such broad questions at the news group,
    > that you do some preliminary research and then you can raise more focused
    > issues.
    >
    > earl
    >
    >
    > --
    > Earl Cox
    > VP, Research/Chief Scientist
    > Panacya, Inc.
    > 134 National Business Parkway
    > Annapolis Junction, MD 20701
    > (410) 904-8741
    > -------------------------------------------
    >
    > AUTHOR:
    > "The Fuzzy Systems Handbook" (1994)
    > "Fuzzy Logic for Business and Industry" (1995)
    > "Beyond Humanity: CyberEvolution and Future Minds"
    > (1996, with Greg Paul, Paleontologist/Artist)
    > "The Fuzzy Systems Handbook, 2nd Ed." (1998)
    > "Fuzzy Tools for Data Mining and Knowledge Discovery"
    > (due Early Fall, 2001)
    >
    >
    >
    >
    > "default" <default@default.com> wrote in message
    > news:3B787E11.93523FAA@default.com...
    > > Hello Everyone!
    > >
    > > How might a person use fuzzy theory in a database system or inside of an
    > > application?
    > >
    > > Can fuzzy theory be used in a practical manner without using too much
    > > computing time or too many computer resources?
    > >
    > > Is fuzzy theory something that is practical at this time?
    > >
    > > =====
    > > Fractal A.
    > > fractala@yahoo.com
    > >
    > >
    >
    >
    >
    > ############################################################################
    > 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 listproc@dbai.tuwien.ac.at
    > (2) To unsubscribe from this mailing list, send a message body of
    > "UNSUB FUZZY-MAIL" or "UNSUB FUZZY-MAIL yoursubscription@email.address.com"
    > to listproc@dbai.tuwien.ac.at
    > (3) To reach the human who maintains the list, send mail to
    > fuzzy-owner@dbai.tuwien.ac.at
    > (4) WWW access and other information on Fuzzy Sets and Logic see
    > http://www.dbai.tuwien.ac.at/ftp/mlowner/fuzzy-mail.info
    > (5) WWW archive: http://www.dbai.tuwien.ac.at/marchives/fuzzy-mail/index.html
    >

    ############################################################################
    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 listproc@dbai.tuwien.ac.at
    (2) To unsubscribe from this mailing list, send a message body of
    "UNSUB FUZZY-MAIL" or "UNSUB FUZZY-MAIL yoursubscription@email.address.com"
    to listproc@dbai.tuwien.ac.at
    (3) To reach the human who maintains the list, send mail to
    fuzzy-owner@dbai.tuwien.ac.at
    (4) WWW access and other information on Fuzzy Sets and Logic see
    http://www.dbai.tuwien.ac.at/ftp/mlowner/fuzzy-mail.info
    (5) WWW archive: http://www.dbai.tuwien.ac.at/marchives/fuzzy-mail/index.html



    This archive was generated by hypermail 2b30 : Wed Aug 15 2001 - 13:42:11 MET DST