Genetic Algorithms, Fuzzy Logic

1
Genetic Algorithms Fuzzy Logic Intelligent
Agents

• Presented By
• Dog Gone I.T.
• Lance Peiper
• Liz Cooksey
• Josh Berwald
• Eric Rowe
• Jill McClain

2
What is an Algorithm

• A set of instructions that is repeated to solve a
  problem.

3
What are Genetic Algorithms

• Software programs that learn in an evolutionary
  manner similarly to the way biological systems
  evolve.
• Simply it is a search method that follows a
  process that stimulates evolution in a computer.
• Survival of the Fittest solution.

4
Process of a Genetic Algorithm
Start
Describe Problem
Generate Solutions
Stop
Yes
Is it good enough
Step 1 Step 2 Step 3 Step 4 Step 5
NO
Select best parents to reproduce
Create a set of offspring
5
First Step

• When given a problem
• Describe it
• Generate a Solution - represented as a
  chromosome (string of genes)
• Test the Solution
• If it is the best then STOP.
• If it is not good enough go to Step 2..

6
Step 2

• Reproduction
• Select parents (solutions) with a max. fitness
  function to produce offspring
• Fitness Function - measure of the objective to be
  obtained
• Parents can also be chosen by random selection
• Result - a new generation of solutions

7
Step 3

• Crossover Process
• Choosing a random position in the string and
  exchanging the segments by either moving it left
  or right with another string

Step 4

• Creation of new offspring

8
Step 5

• Apply random mutation
• an arbitrary change in a situation (string)
• used to prevent the algorithm from getting
  stuck

Repeat to Step 1
9
Genetic Algorithm Example

• Secret Number 001010
• Random Trial and error 32 guesses (avg)
• Genetic Algorithm Solution
• Guess 4 numbers (each candidate solution is
  called a chromosome)
• (A) 110100 1 digit correct Delete
• (B) 111101 1 digit correct Delete
• (C) 011011 4 digits correct
• (D) 101100 3 digits correct
• For C D Mate the parents genes by splitting
  each number
• C) 011011 D) 101100
• Crossover to create offsprings combine 1st part
  of C with 2nd part of D and vice versa
• E) 011100 - score 3 F)101011 - score 4 (new
  generation)
• Offsprings not much better than parents. Mate
  and crossover parents with new split
• C) 011011 D) 101100
• G) 011000 score 4 H) 101111 Score 3
• Select best couple out of C H. Number of
  different options (eg. CF or CG GF). Mate
  and crossover
• F) 101011 G) 011000 I) 111000 score 3
  J) 001011 score 5
• F) 101011 G) 011000 k) 101000 score
  4 L) 011011 score 4
• Mate and crossover with J and K as parents

10
Parameters to Set

• Number of initial solutions to generate
• Number of offspring to generate
• Number of parents and offspring to keep for the
  next generation
• Mutation probability (very low)
• Probability distribution of crossover point
  occurrence

11
Parameters to Set (cont.)

• Parameters depend on problem being solved
• Determined by trial and error

12
GA is used in..

• Dynamic Process Control
• Simulation of models of behavior and evolution
• Complex design of engineering structures

• Pattern Recognition
• Scheduling
• Transportation and Routing
• Layout and Circuit design
• Telecommunications

13
GA Software

• Evolver (Excel add-in)
• Genetic Algorithm User Interface (Adaptive
  Software www.gaui.com)
• XperRule GenAsys - an ES shell with an embedded
  genetic algorithm

14
Fuzzy Logic

• Uncertainty
• Accurate way to describe human decision making
• Thinking isnt always true or false
• Involves the gray areas
• Allows computer to be less precise

15
Examples of Fuzzy Logic

• Anti-lock brakes may lock by sensing 18 different
  factors
• Fuzzy logic washing machines use three inputs and
  sensors to adjust the load parameters and time to
  wash and rinse
• Fuzzy logic vacuums monitor the surface and the
  condition of the material and adjust
  automatically.

16
Advantages

• Provides flexibility
• Provides options
• Allows for observation
• Increases the systems maintainability
• Controls situation not easily defined by
  mathematical solutions

17
Intelligent Agent (IA)

• A computer program that helps a user with routine
  computer tasks
• Relatively new technology
• Several definitions and fluctuating capabilities
• Overcome information overflow on the internet

18
Intelligent agents

• Assists user with tasks such as searching
• agents decide what is relevant to user
• Filters data identifies relevant sources of data
• Agent is autonomous
• Goal oriented
• Flexible
• Self-starting

19
Autonomy

• Example- Science Experiment
• User Specifies
• Needs- find journal articles about experiments in
  a specific field
• Constraints- a certain experimental system or
  reactant/reagent was used
• And preferences- articles are written by research
  groups at major universities
• Agent then makes various decisions based on
  parameters set- they dont need constatn user
  input.

20
Autonomy

• Autonomy provides time saving capabilities
  because it works for the user when the user isnt
  directly in control. Too much autonomy can be
  problematic. A good agent needs to find the right
  level of autonomy for the task at hand.

21
IA applications

• Building communities
• Match people based on preferences
• Use criteria such as musical tastes books
  investment analysis etc.
• Agents inc

22
Research analysis news

• Reduces information overload
• Saves search time
• Who has most/best content
• Altavista- advanced searching
• Farcast- custom news
• IBM infomarket

23
Products and services

• Value added tools
• Finds products and services from company
  databases
• Returns product descriptions and prices
• Vendors pay to be included
• Fido the shopping doggie

24
Evolver
Evolver is an optimization add-in for Microsoft
Excel. Evolver uses innovative genetic algorithm
(GA) technology to quickly solve complex
optimization problems in finance distribution
scheduling resource allocation manufacturing
budgeting engineering and more.
http//www.palisade.com/html/evolver.html
25
Tutorial

• Presented By
• Dog Gone I.T.
• Lance Peiper
• Liz Cooksey
• Josh Berwald
• Eric Rowe
• Jill McClain

26
First Things First

• Download Evolver 4.0 Demo from
• http//www20.brinkster.com/mist5620
• Also download tutorials and tutorial excel
  files
• Install Evolver and extract tutorial files
• Start Programs Palisade Decision Tools
  Evolver 4.0 for Excel

27
Display Watcher
Start Optimization
Start Evolver
Evolver Settings
Stop Optimization
28
Evolver Settings

• Select the target cell the cell that you are
  trying to maximize minimize or get close to a
  particular value.
• Add adjustable cells the cells that Evolver
  will adjust to try and find the optimal solution.
• Define constraints constraints are conditions
  that must be met for a solution to be valid.

29
Target Cell
Add Adjusting Cells
Add Constraints
Options Dialog
30
Add Adjusting Cells
Choose Solving Method
Adjustable Cell Range
Minimum Cell Value
Maximum Cell Value
Allow Only Integers
31
Solving Methods

• Recipe variables can be combined independently
  like ingredients in a recipe.
• Order used when you are trying to arrange a set
  of given values into a particular order.
• Grouping variables are combined into groups
  where the number of groups equals the number of
  unique numbers.
• Budget similar to the Recipe method except that
  you are shooting for a summed target goal.
• Project used mainly for project management
  where items must be completed in a particular
  order.
• Schedule similar to the Group method but
  assigns each variable to a unit of time.

32
Adding Constraints

• Hard Constraints conditions that must be met
  for a solution to be valid. Solutions that dont
  meet the constraint will be thrown out.
• Soft Constraints conditions that we would like
  to meet but the solutions are not thrown out it
  the condition is not met.

33
Evolver Options
Updating the display
Graph Progress Option
Stop after so many trials or after so many minutes
34
Dog-Gone I.T.

• Thanks!
• Any questions