1. optimization and its necessity. classes of optimizations problems. evolutionary optimization....

22
Optimization A focus on evolutionary optimization and its applications Daniel Khashabi ([email protected]) Amirkabir University of Technology, School of Electrical Engineering October 20, 2010 Introduction to 1

Upload: duane-nash

Post on 26-Dec-2015

215 views

Category:

Documents


2 download

TRANSCRIPT

Page 1: 1. Optimization and its necessity. Classes of optimizations problems. Evolutionary optimization. –Historical overview. –How it works?! Several Applications

OptimizationA focus on evolutionary

optimization and its applications

Daniel Khashabi ([email protected])Amirkabir University of Technology, School of Electrical

EngineeringOctober 20, 2010

Introduction to

1

Page 2: 1. Optimization and its necessity. Classes of optimizations problems. Evolutionary optimization. –Historical overview. –How it works?! Several Applications

Lecture Overview:• Optimization and its necessity.• Classes of optimizations problems.• Evolutionary optimization.

– Historical overview.– How it works?!

• Several Applications of EO.– Examples.

2

Page 3: 1. Optimization and its necessity. Classes of optimizations problems. Evolutionary optimization. –Historical overview. –How it works?! Several Applications

OptimizationA simple function: - Remember derivation in math(I) course! - The goal: finding maximum and minimum - Best answer: Global max/min

General Form Definition: • Find set which maximizes function

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

A

B

C

D

E

F

G

f'(x)=0f"(x)<0

f'(x)=0f"(x)>0

3

Page 4: 1. Optimization and its necessity. Classes of optimizations problems. Evolutionary optimization. –Historical overview. –How it works?! Several Applications

Local vs. Global; a BIG challenge!

• This an important challenge !

[Optimization with Genetic Algorithm/Direct Search Toolbox : Ed Hall]

1 2 3( , , ,..., )nx x x x

4

Page 5: 1. Optimization and its necessity. Classes of optimizations problems. Evolutionary optimization. –Historical overview. –How it works?! Several Applications

Necessity of OptimizationEvery engineering design can be assumed as a black-box :

e.g. a robot, an antenna, a machine, a network, a program , …

Aim is to design black-box with • enough performance• least cost! Optimization !

5

Page 6: 1. Optimization and its necessity. Classes of optimizations problems. Evolutionary optimization. –Historical overview. –How it works?! Several Applications

Necessity of OptimizationSome engineering design examples: Analog Filter design: Goal: to find a minimal arrangement of elements which gives us desired frequency response!Elements: • Self inductor • Capacitor• Resistor• ...

Parameters: • Arrangement of elements makes the frequency response.

6

Page 7: 1. Optimization and its necessity. Classes of optimizations problems. Evolutionary optimization. –Historical overview. –How it works?! Several Applications

Necessity of Optimization

Some engineering design examples: Electrical machine design:Goal: design a motor which has best performance(Low loss)How? • Changing internal structure of a motor(say dc motor)

Performance should be modeled As a function!

Elements:• Number of commutator• Direction/number of

compensating windings • …

-> Design parameters

7

Page 8: 1. Optimization and its necessity. Classes of optimizations problems. Evolutionary optimization. –Historical overview. –How it works?! Several Applications

Necessity of OptimizationEvery engineering design needs to be optimized!This is the world of optimization:- Electrical machine design- Robotics- Circuit design- Antenna design- Telecommunication Routing- ….

Other fields:- Structure design e.g.

- Automotive design:

8

Page 9: 1. Optimization and its necessity. Classes of optimizations problems. Evolutionary optimization. –Historical overview. –How it works?! Several Applications

Optimization MethodsThere are lots of optimization methods:

- Gradient Methods.- Linear Programming.- Quadratic Programming.- …- Evolutionary Methods!

• key that specifies which “method of optimization” is suitable for our challenge is characteristics of problem, i.e. complexity of problem:– Number of variables.– Constraints of variables.– Structure of function: Linearity, Quadratic or completely non-

linear.– Derivability of function.– …

1 2 3( , , ,..., )nx x x x1 2 3( , , ,..., )nx x x x1 2 3( , , ,..., )nx x x x1 2 3( , , ,..., )nx x x x

9

Page 10: 1. Optimization and its necessity. Classes of optimizations problems. Evolutionary optimization. –Historical overview. –How it works?! Several Applications

EO: Historical Overview• Inspired from Darwin's “Evolution Theory”.

– Evolution of human generation during time by mutation and crossover(breeding)

– Betters(Fitter) have more chance to survive– This causes generations tend to better characteristics!

• Evolutionary Optimization/Genetic algorithms– Rapidly growing area of artificial intelligence.– Evolves solutions!

[Charles Darwin: 1809-1882 : http://en.wikipedia.org/wiki/Charles_Darwin]

[http://daily.swarthmore.edu/static/uploads/by_date/2009/02/19/evolution.jpg]

10

Page 11: 1. Optimization and its necessity. Classes of optimizations problems. Evolutionary optimization. –Historical overview. –How it works?! Several Applications

Evolutionary Optimization• A way to employ evolution in solutions• Optimization

– Based of variation and selection– by understanding the adaptive processes of natural systems

• Search for ?! – Find a better solution to a problem in a large space.

• What is a better solution? – A good solution is specified by “Fitness Function”!– A “Fitness Function” is a function that shows how answers are desirable !

• E.g. performance of a machine, gain of a circuit, ….

[http://science.kukuchew.com/wp-content/uploads/2008/05/explosm-evolution-t-shirt.jpg]

11

Page 12: 1. Optimization and its necessity. Classes of optimizations problems. Evolutionary optimization. –Historical overview. –How it works?! Several Applications

EO: How it works? • Solution of problem is formed by -> “Population” • Population consists of -> individuals.• Every population is parent generation for next generation.• Solutions are evolved in every generation. How?!

– Crossover and mutation• Individuals that are more fitter -> more chance to survive! • Fitness in population grows gradually, as generations pass.

– This is called “Evolution”!

[“Evolutionary Algorithms”: S.N.Razavi]

12

Page 13: 1. Optimization and its necessity. Classes of optimizations problems. Evolutionary optimization. –Historical overview. –How it works?! Several Applications

Traveling Salesman Problem(TSP)

• A single salesman travels to cities and completes the route by returning to the city he started from.• Each city is visited by the salesman exactly once.• Find a sequence of cities with a minimal travelled distance.

Encoding: Chromosome describes the order of cities, in which the salesman will visit them

4238352621353273846445860697678716967628494

0

20

40

60

80

100

120

0 10 20 30 40 50 60 70 80 90 100

y

x

TSP30 Solution (Performance = 420)

[Genetic Algorithms: A Tutorial: W.Wliliams][http://www.informatik.uni-leipzig.de/~meiler/

Schuelerseiten.dir/TBlaszkiewitz/GermanyLRoute.jpg]

13

Page 14: 1. Optimization and its necessity. Classes of optimizations problems. Evolutionary optimization. –Historical overview. –How it works?! Several Applications

Traveling Salesman Problem(TSP)

14

Page 15: 1. Optimization and its necessity. Classes of optimizations problems. Evolutionary optimization. –Historical overview. –How it works?! Several Applications

Evolvable Hardware

[“Design and Optimizing Digital Combinational Gates”: M.Moosavi, D.Khashabi]

• How to Evolve a Hardware ?! “Design and Optimizing a digital combinational logic circuit using GA.”

• Example Run:

15

Page 16: 1. Optimization and its necessity. Classes of optimizations problems. Evolutionary optimization. –Historical overview. –How it works?! Several Applications

Which one is better?!

Evolving a Bicycle!

16

Page 17: 1. Optimization and its necessity. Classes of optimizations problems. Evolutionary optimization. –Historical overview. –How it works?! Several Applications

Goal: evolves a machine that is able to traverse most distance!Parameters: • Wheel and mass diameter• Springs length and stiffness

Evolving a Bicycle!

17

Page 18: 1. Optimization and its necessity. Classes of optimizations problems. Evolutionary optimization. –Historical overview. –How it works?! Several Applications

• Control – Gas pipeline, pole balancing, Robot motion

planning and obstacle avoidance … • Design Problems

– Semiconductor Design, Aircraft Design, Keyboard configuration, Resource Allocation(e.g. electrical power networks.)

• Signal Processing: – Filter design

• Automatic Programming– Genetic Programming…

Applications of Evolutionary Optimization in a nutshell !

18

Page 19: 1. Optimization and its necessity. Classes of optimizations problems. Evolutionary optimization. –Historical overview. –How it works?! Several Applications

Use MATLAB!• Optimization Toolbox:

optimtool• Genetic Algorithm Toolbox:

gatool

19

Page 20: 1. Optimization and its necessity. Classes of optimizations problems. Evolutionary optimization. –Historical overview. –How it works?! Several Applications

• Optimization and …– its necessity

• Evolutionary optimization– Historical foundation– Procedure

• Several examples and applications.

Summery

20

Page 21: 1. Optimization and its necessity. Classes of optimizations problems. Evolutionary optimization. –Historical overview. –How it works?! Several Applications

Question?

Thanks!

21

Page 22: 1. Optimization and its necessity. Classes of optimizations problems. Evolutionary optimization. –Historical overview. –How it works?! Several Applications

References:

• [1] Wikipedia.com• [2] K.Kiani, Presentation: “Genetic Algorithms” .• [3] W.Wliliams, Presentation: “Genetic Algorithms:A

Tutorial”.• [4] S.N.Razavi, Presentation: “Evolutionary Algorithms”.• [5] M.Moosavi, D.Khashabi, “Designing and Optimizing

Digital Combinational Logic Circuits”, Iranian Student Conference of Electrical Engineering, August-2010.

22