Scribd
Upload
33 views

Optimization Heuristics: Kausal Malladi

This presentation discusses various optimization heuristics including genetic algorithms, hill climbing, tabu search, and simulated annealing. It provides an overview of each technique and examples of their application to problems like the travelling salesman problem. Genetic algorithms mimic natural evolution to evolve solution populations toward better solutions over generations. Hill climbing iteratively moves to better solutions incrementally but may get stuck at local optima. Tabu search prohibits visiting recently considered solutions to avoid cycles and overcome local optima. Simulated annealing is inspired by annealing in metallurgy and gradually accepts worse solutions based on temperature to approximate global optima. Swarm intelligence techniques like ant colony optimization also inform optimization heuristics.

Uploaded by

Nazir Kazimi
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
33 views

Optimization Heuristics: Kausal Malladi

This presentation discusses various optimization heuristics including genetic algorithms, hill climbing, tabu search, and simulated annealing. It provides an overview of each technique and examples of their application to problems like the travelling salesman problem. Genetic algorithms mimic natural evolution to evolve solution populations toward better solutions over generations. Hill climbing iteratively moves to better solutions incrementally but may get stuck at local optima. Tabu search prohibits visiting recently considered solutions to avoid cycles and overcome local optima. Simulated annealing is inspired by annealing in metallurgy and gradually accepts worse solutions based on temperature to approximate global optima. Swarm intelligence techniques like ant colony optimization also inform optimization heuristics.

Uploaded by

Nazir Kazimi
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 17

A presentation on

Optimization Heuristics
by
Kausal Malladi
(Student, IIIT Bangalore)

Agenda

Definition of a Heuristic

Optimization heuristics

Genetic Algoriths

!ill "libing

Tabu Search

Siulated Annealing

S#ar Intelligence

$ith e%aple applications



Heuristic

&%perience based techni'ues for proble


sol(ing, learning and disco(ery) *Adopted fro
$i+ipedia,

Different types

-ule of thub

"oon sense

&ducated Guess

Meta.heuristics/ 0araeters that influence


eploying a heuristic

Optimization Heuristics

Al#ays difficult to sol(e NP-Hard and NP-


Complete coputational probles

&(en #ith different optii1ation techni'ues,


actual running tie is ne(er guaranteed

$e eploy soe rules 2 results based on


e%perients to state that a near.optial
solution can be obtained

3o proof as to #hy and ho# #e get


solution

Genetic Algorithms

A heuristic that iics natural evolution

A population of Candidate Solutions


e(ol(ed to#ards better solutions

Generations

-e'uires

Genetic Representation of solution doain

Fitness function to e(aluate solution

Applications/ Gae Theory



Local Search

To sol(e hard 4ptii1ation probles

Search Space / Doain of function to be


optii1ed

5inding a solution aong nuber of


candidate solutions, a%ii1ing a criterion

Sub.failies/

!ill "libing

Tabu Search

Siulated Annealing

Hill Climbing

Iterative algorith, starts #ith arbitrary


solution

6oo+s for better solutions increentally

-epeats until no further ipro(eents

Good for finding a local optimum

Doesn7t guarantee global optimum

Siple, popular

$or+s #ell, generally



Hill Climbing

0opular e%aple 8 TS0

Tra(elling Salesan 0roble

Kno#n NP-Hard proble

Initial solution ay not be optial

Shorter route is ore li+ely to be obtained

$idely used in Artificial Intelligence

Significant results in real.tie systes

Any-time algorith

0itfall/ 0lateau

Tabu Search

Iterati(ely proceeds fro one potential


solution S to an ipro(ed one S in the
neighbourhood of S

4(ercoes fe# pitfalls of other Local


Search techni'ues (&%aple/ Plateau)

9isited solutions ar+ed !tabu"

Search progresses using #emor$


Structures

4ften a benchar+ heuristic:



Tabu Search

Meory structures

Describe

9isited solutions

;ser pro(ided sets of rules

"ategories

Short ter

Interediate ter

6ong ter

5or tabu list



Tabu Search

Issues

4nly effecti(e in discrete search spaces

$or+around/ A similarity measure

!igh diensional search space

$or+around/ "reate a tabu list consisting


of attributes of a solution

"an be ore effecti(e solution, has


probles too

Aspiration criteria introduced

4(erride solution7s tabu state



Tabu Search

"oon e%aple 8 TS0

Tra(elling Salesan 0roble

Tabu Search finds a satisficing solution

Starts #ith an initial solution that can be


found randoly or using soe algorith

4rder in #hich t#o cities are (isited, is


s#apped

Total tra(elling distance is the etric

A acceptable solution added to tabu list if


in neighbourhood of accepted solution

Simulated Annealing

Inspiration/ Annealing in Metallurgy

0robabilistic eta.heuristic

Appro%iates global optimum in a large


search space

Gi(es acceptably good solution if not the


best

Slo# decrease in probability of accepting


#orse solutions

Simulated Annealing

&%aple 8 TS0

Tra(elling Salesan 0roble

Metric under consideration is Mileage

#etropolis Algorithm

0air#ise changing order of (isit to cities

Solutions that don7t lo#er ileage also


accepted

e
.D/T
> R(0,1)

D is the change of distance iplied

If T is large, any bad choices are ade



S%arm &ntelligence

A collecti(e beha(ior of self.organi1ed


systes #hich are decentrali1ed *Adopted fro
$i+ipedia,

"an7t predict ho# the systes beha(e


e(en #ithout a centrali1ed control

$idely eployed in Artiicial Intelligence

&%aple

Ant "olony 4ptii1ation

3atural ants Siulation agents

0heroones -ecording position, 'uality



'e(erences

http/22###)iaeng)org2publication2$"&<==>2$"&<==>?pp@A.@B)pdf
(Gae Theory using Genetic Algoriths)

http/22ath#orld)#olfra)co2SiulatedAnnealing)htl
(Siulated Annealing)

http/22artificialintelligence.notes)blogspot)in2<=A=2=>2hill.clibing.procedure)htl
(!ill "libing in Artificial Intelligence)

Than+ you:

You might also like