Re: ANN: Fuzzy sets for Ada v4.1

["Dmitry A. Kazakov" <mailbox@dmitry-kazakov.de>]

On Tue, 13 Sep 2005 08:34:42 -0500, Marc A. Criley wrote:

> Dmitry A. Kazakov wrote:
>> It is here:
>> 
>> http://www.dmitry-kazakov.de/ada/fuzzy.htm
> 
> Just curious...
> 
> It's been several years since I looked at fuzzy logic--just one of those 
> things I was never able to get back to :-)
> 
> What are some of the application areas that you or others are using your 
> fuzzy sets for?

I'm using it for machine learning and pattern recognition. Examples:
classification of vessels by acoustic signals; a production quality control
system based on video images; a customer support system for ordering
components needed to assemble a test bench.

In general "fuzzy" = uncertain, but not like "random", it is worse, more
uncertain. Usually human inputs are fuzzy. Also models can be fuzzy, when
some parameters are unknown (and cannot be statistically estimated.) In
fact it is again a kind of human [expert] input. In many cases uncertainty
is mixed: randomness x fuzziness.

The problem with applications of fuzzy is that historically many people
believed fuzzy to be a universal theory to replace and explain everything.
It was wrong. Moreover, for any problem the rule of thumb is: first to try
to formulate it as deterministic certain. If that does not work, then as
stochastic random. And only after that as fuzzy. So as long as your methods
work you need not to come back! (:-))

-- 
Regards,
Dmitry A. Kazakov
http://www.dmitry-kazakov.de