# Math Lecture Series Fuzzy Nueral Networks

AI and Robotics

Oct 20, 2013 (4 years and 8 months ago)

94 views

Lukas Yanni

Professor Porter

Calculus III

12/3
/2012

Fuzzy Neural
Networks

and You

This presentation was particularly cool for me, as my current job in the network
department of the school. I work with router, switch panels, and other informational
systems
technologies. Thus, I was naturally interested in learning more about different types of networks
and how they can be improved and interact with each other. This talk defined what fuzzy logic
systems and neural networks are. His main premise was ho

each of them excels and their
drawbacks are, and how they can be improved by being combined into fuzzy neural networks.
His presentation used many good examples to display each, and then how they could be
implemented together in a practical way.

His desc
ription of a fuzzy logic system was quite clear. He explained that a it basically is
a system of parameters that a detected situation would fall into, such a black, dark grey, light
grey, and white, and for each scenario an output function would be perform
ed. These work for
simple situations quite nicely, but once the scenarios become increasingly complicated,
the
system works less well, as it is not able to adapt to a dynamic environment once the situation no
longer fits the parameters established.

A neu
ral network is based off of how the human brain functions. It is a “plastic” system
that can le
arn, recognize patterns, and

adapt based on how the network is being used and what
input it sees on a regular basis. Such networks are used in facial recognition

software. It does this
by sensing a pattern in images fed into its system, and sensing a unique type of pattern in the
picture, which the user identifies as a face. Neural networks
are able to adapt, be corrected, can
be generalize, and can solve complex
problems without a mathematical model. However it is
hard for users to see the mechanisms behind the problem solving. It cannot give a formula, and is
difficult to test.

Combining fuzzy logic and neural networks allows neural networks to learn and
optimi
ze, are easier to analyze and modify, and provide the best of both.
Overall, it seems like a
good combination and will be involved in the future of certain network technology.