Re: The relationship btw Fuzzy & Neural Net?

Mark A. Scarton (marksc@wpmail.code3.com)
Fri, 29 Mar 1996 16:01:48 +0100


wilson@ohm.nii.ncb.gov.sg (Wilson Tam Siu Hom) wrote:
>When I ask some body about what is Fuzzy logic, more or less they will try
>to refer to Neural Network. However, I don't really understand the
>relationship between Fuzzy and Neural Net.... Can anyone help me?

Fuzzy is mostly used when you have a system with uncertainty factors in the data
.. ambiguity, imprecision, discord (conflict), etc. Neural networks are mostly
used when you have a system about which you have uncertainty about the inner
relationships among system components (subsystems).

So if the data is imprecise (or you are willing to make the data imprecise to
lower the complexity modelling the system in question), fuzzy gives you tools
for building the system. You know how the system is organized, the
relationships among the data variables, but the values themselves are fuzzy.

If the data values are precise, but you really don't know how they relate, try
establishing a neural network model and see if it can be trained to recreate the
behavior of the system being modelled.

And of most recent interest, some complex systems include both factors. Hence
the interest in fuzzified neural networks.

You might contact the BISC group at Berkeley for references and/or additional
information. Their www page is at "http://www.cs.berkeley.edu/projects/bisc/".
They deal with the intertwining of soft computing approaches.

Mark A. Scarton, ABD
CompUtah!, Park City, Utah USA
Home: 801.565.9835
Office: 801.265-4612