ENDORSING PARTNERS

The roles of evolutionary
computation, fitness landscape,
constructive methods and local searches
in t...
The Roles of Evolutionary Computation,
Fitness Landscape, Constructive Methods
and Local Searches in the Development of
Ad...
The roles of evolutionary computation, fitness landscape,
constructive methods and local searches in the development
of ad...
Outline
• Introduction

– Infrastructure problems.
• Simple representative problems.
• Hybrid model description.
• Benchma...
Infrastructure
• Connected elements which are structurally related and
each element affects other elements.
• Involved wit...
Z

Infrastructure problems difficulty

(0,0,0)

X

Number of Variables

Probabilistic Nature
Ti
m
eH
or
iz
on

(1,1,1)

Y
Simplified problem: minimizing traffic
Library(1)

A

Avg. traffic

Hospital (2)

distance

University (5)

C

B

E

D

A
...
The complexity of the problems

𝑛
𝑛
Traffic_Volume(𝜋) = ∑ 𝑖=1 ∑ 𝑗=1 𝑓𝑖𝑖 𝑑 𝜋 𝑖 𝜋(𝑗)
The problem is called Quadratic Assignm...
Speed of computers
• This “intractability” is despite the fact that the computer
industry has progressed so fast.
“If the ...
The proposed hybrid meta-heuristic
• The aim is to develop effective
problem-solving procedures to obtain
high quality sol...
Fitness landscape
• When solved with local search methods, the difficulty of an
optimization problem is directly related t...
Modular design
• Based on “No-free-lunch” theorem
(Wolpert and Macready 1997),
incorporating problem-specific
knowledge, a...
Benchmark test results
• Designed Algorithms have been applied to three wellknown problems in the literature:

– Quadratic...
References
Boese, K. D., A. B. Kahng, et al. (1994). "A new adaptive multi-start technique for
combinatorial global optimi...
Mehrdad Amirghasemi
PhD student
ma604 (at) uow.edu.au

15
Upcoming SlideShare
Loading in …5
×

SMART International Symposium for Next Generation Infrastructure: The roles of evolutionary computation, fitness landscape, constructive methods and local searches in the development of adaptive systems for infrastructure planning

664 views

Published on

A presentation conducted by Mr Mehrdad Amirghasemi, SMART Infrastructure Facility, University
of Wollongong.
Presented on Wednesday the 2nd of October 2013.

Modelling and simulation for improved infrastructure is involved with the development of adaptive systems that can learn and respond to the environment intelligently. Developing simple agents with limited intelligence that collectively represent complex behaviour can assist infrastructure planning and can model many real world
situations. By employing sophisticated techniques which highly support infrastructure planning and design, evolutionary computation can play a key role in the development of such systems. The key to presenting solution strategies for these systems is fitness landscape
which makes some problems hard and some problems easy to tackle. Moreover, constructive methods and local searches can assist evolutionary searches to improve
their performance. In this paper, all these four concepts are reviewed and their application in infrastructure planning and design is discussed. With respect to applications, the main emphasis includes city planning, and traffic equilibrium.

Published in: Education, Technology
0 Comments
1 Like
Statistics
Notes
  • Be the first to comment

No Downloads
Views
Total views
664
On SlideShare
0
From Embeds
0
Number of Embeds
1
Actions
Shares
0
Downloads
5
Comments
0
Likes
1
Embeds 0
No embeds

No notes for slide

SMART International Symposium for Next Generation Infrastructure: The roles of evolutionary computation, fitness landscape, constructive methods and local searches in the development of adaptive systems for infrastructure planning

  1. 1. ENDORSING PARTNERS The roles of evolutionary computation, fitness landscape, constructive methods and local searches in the development of adaptive systems Monday, 30 September 2013: Business & policy Dialogue for infrastructure planning The following are confirmed contributors to the business and policy dialogue in Sydney: • Rick Sawers (National Australia Bank) • Nick Greiner (Chairman (Infrastructure NSW) th Tuesday 1 October to Thursday, Dialogue 3rd October: Academic and Policy www.isngi.org Presented by: Mr Mehrdad Amirghasemi, SMART Infrastructure Facility, University of Wollongong www.isngi.org
  2. 2. The Roles of Evolutionary Computation, Fitness Landscape, Constructive Methods and Local Searches in the Development of Adaptive Systems for Infrastructure Planning Mehrdad Amirghasemi Reza Zamani
  3. 3. The roles of evolutionary computation, fitness landscape, constructive methods and local searches in the development of adaptive systems for infrastructure planning
  4. 4. Outline • Introduction – Infrastructure problems. • Simple representative problems. • Hybrid model description. • Benchmark test results.
  5. 5. Infrastructure • Connected elements which are structurally related and each element affects other elements. • Involved with the optimized selection of a set of values, among a set of alternatives, for a number of variables. • Interaction among variables makes these problems extremely hard. • Sophisticated optimization techniques and mathematical programming are the key issues in reaching a reasonable level of efficiency in solving infrastructure problems.
  6. 6. Z Infrastructure problems difficulty (0,0,0) X Number of Variables Probabilistic Nature Ti m eH or iz on (1,1,1) Y
  7. 7. Simplified problem: minimizing traffic Library(1) A Avg. traffic Hospital (2) distance University (5) C B E D A B C D E A B C D E 1 4 2 3 π1 5 5 1 3 2 π2 4 Shopping Centre (4) Sport Centre (3)
  8. 8. The complexity of the problems 𝑛 𝑛 Traffic_Volume(𝜋) = ∑ 𝑖=1 ∑ 𝑗=1 𝑓𝑖𝑖 𝑑 𝜋 𝑖 𝜋(𝑗) The problem is called Quadratic Assignment Problem(QAP). As the problem size increases linearly, the size of solution space increases exponentially. 5!=5×4×3×2×1=120 60! = 60×59× … ×2×1 = 8×1081 With an 8 Ghz processor, evaluating all possibilities takes 1066 years of computation time. The age of the universe is around 1010 years. • Simplified mathematical model: • • • • •
  9. 9. Speed of computers • This “intractability” is despite the fact that the computer industry has progressed so fast. “If the car industry moved as fast as the computer industry, cars would get 470, 000 mph, 100,000 miles per gallon, and would cost three cents.” –Paul Otellini, Intel CEO. • This necessitates a need for better algorithms that are capable of yielding high-quality solutions in a reasonable amount of time.
  10. 10. The proposed hybrid meta-heuristic • The aim is to develop effective problem-solving procedures to obtain high quality solutions fast. • Generally the designed algorithms consist of four modules: – A constructive procedure to produce a pool of high quality initial solutions. – A Local Search procedure to improve a given, complete solution. – A population based (evolutionary) module to combine solutions from the current pool. – A Synchronizer module, which facilitate interaction among the above three modules. Constructive Method Synchronizer Local Search Genetic Algotihm
  11. 11. Fitness landscape • When solved with local search methods, the difficulty of an optimization problem is directly related to the shape of its fitness landscape. The fitness landscape of a hard optimization problem The fitness landscape of an easy optimization problem
  12. 12. Modular design • Based on “No-free-lunch” theorem (Wolpert and Macready 1997), incorporating problem-specific knowledge, and matching the “procedure” with the “problem” is essential for developing high performance procedures. • The ideal case is to have a generic procedure to handle similar problems. • A modular design helps to achieve a balance between the above two facts. 12
  13. 13. Benchmark test results • Designed Algorithms have been applied to three wellknown problems in the literature: – Quadratic Assignment, QAP – Job Shop Problem, JSP – Permutation Flow Shop Problem, PFSP • Competitive results have been achieved on all benchmark tests. • For JSP, a notorious instance is solved 6 times faster the fastest available method in the literature.
  14. 14. References Boese, K. D., A. B. Kahng, et al. (1994). "A new adaptive multi-start technique for combinatorial global optimizations." Operations Research Letters 16(2): 101-113. De Jong, K. A. (2006). Evolutionary computation: a unified approach, MIT Press. Fogel, D. B. (1994). "An introduction to simulated evolutionary optimization." Neural Networks, IEEE Transactions on 5(1): 3-14. Jones, T. and S. Forrest (1995). Fitness Distance Correlation as a Measure of Problem Difficulty for Genetic Algorithms. Proceedings of the 6th International Conference on Genetic Algorithms, Morgan Kaufmann Publishers Inc.: 184-192. Wolpert, D. H. and W. G. Macready (1997). "No free lunch theorems for optimization." Evolutionary Computation, IEEE Transactions on 1(1): 67-82. http://www.iconarchive.com/ http://maps.google.com/
  15. 15. Mehrdad Amirghasemi PhD student ma604 (at) uow.edu.au 15

×