# Genetic Algorithms -- System

AI and Robotics

Oct 23, 2013 (4 years and 6 months ago)

96 views

3/21/02

Genetic Algorithms
--

System
Modeling and Function Finding

1

Genetic Algorithms
--

System
Modeling and Function Finding

Jack Perdue

CPSC 689
-
608

March 21, 2002

3/21/02

Genetic Algorithms
--

System
Modeling and Function Finding

2

Problem Statement

Given a set of observed metrics
(variables) where some metrics are
dependent upon others (but the
relationship is unclear), develop a
function/model for the the
dependent metrics.

3/21/02

Genetic Algorithms
--

System
Modeling and Function Finding

3

Real
-
World Examples

Predicting freeway speed given the number
of cars entering and exiting, the day of the
week, time of day and weather conditions.

Predicting the winner of a sports event
given statistical history of teams/individuals
involved.

Predicting how long a program will run on a
particular (parallel) computer system given
a history of its past performance and impact
on the system.

3/21/02

Genetic Algorithms
--

System
Modeling and Function Finding

4

A Simple Example

Here we have some
“observed” metrics. It is
instantly apparent that the
function here is something
along the lines of:

x * y = 36 or y = 36 / x

3/21/02

Genetic Algorithms
--

System
Modeling and Function Finding

5

A Harder Example

Here we have some
“observed” metrics. If we
were to plot it, we would see it
is a circle, but it isn’t apparent
just looking at the numbers.

(x
-
1)
2

+ (y
-
2)
2

= 3
2

or

y = 2
±

sqrt(3
2

-

(x
-
1)
2
)

3/21/02

Genetic Algorithms
--

System
Modeling and Function Finding

6

Higher dimensions

Although we can visualize and conceptualize
functions of two or three variables, the human
mind’s ability to deduce patterns is decreased as
the dimensionality is increased.

We have methods such as linear least squares
(for example) to help us with larger dimensions,
but given a large volume of observed metrics,
they often are computationally prohibitive.

3/21/02

Genetic Algorithms
--

System
Modeling and Function Finding

7

Enter GA/GP/GEPs

As we have been learning the past
few weeks, when we have a large
problem space and are trying to find a
needle in a haystack (or something
resembling a needle), then the field of
genetic algorithms and programming
may be of help.

3/21/02

Genetic Algorithms
--

System
Modeling and Function Finding

8

GAs vs. GP vs. GEP

(according to Candida Ferreira)

GAs
-

individuals are linear strings of
fixed length (whole string considered)

GP
-

individuals non
-
linear entities of
different sizes and shapes (parse trees)

GEP
-

individuals are encoded and linear
strings of fixed length but interpreted as
non
-
linear entites of different sizes and
shapes (partial strings considered)

3/21/02

Genetic Algorithms
--

System
Modeling and Function Finding

9

So, what is Gene Expression
Programming (GEP)

... on to Candida Ferreira’s
tutorial...