没有合适的资源?快使用搜索试试~ 我知道了~
资源推荐
资源详情
资源评论
ELEDIAResearchCenter
ELEDIA@UniTN ‐ UniversityofTrento
ViaSommarive 9,I‐38123Trento,Italy
E‐mail:andrea.massa@unitn.it
Web:www.eledia.org/eledia‐unitn
Laboratoire desSignauxetSystèmes
UMR8506(CNRS‐CENTRALESUPELEC‐UNIV.PARISSUD)
rueJoliot‐Curie3,91192Gif‐sur‐Yvette,France
E‐mail:andrea.massa@l2s.centralesupelec.fr
Web:www.eledia.org/eledia‐l2s
Prof.AndreaMASSA
OptimizationTheory,Techniques,
andMATLABsimulations
Global(Evolutionary) OptimizationMethods
UESTC“2018 InternationalSummerSchool”
July16‐20, 2018 – Chengdu,China
2
ELEDIAResearch Center–AllRightsReserved
Allrightstothecontentofthisdocumentarereserved.Anyuse,inwholeor
inpart,ofthecontentsincludedinthisdocument,includingthestorage,
reproduction,editing,disseminationordistributionoftheircontentthroughany
technologyplatform,support,orcomputernetworkisforbiddenwithoutthe
priorwrittenpermissionfrom
theauthor.
Tuttiidirittirelativiaicontenutidelpresentedocumentosonoriservati.È
vietatoqualsiasiutilizzo,totaleoparziale,deicontenutiinseritinelpresente
documento,iviinclusalamemorizzazione,riproduzione,rielaborazione,
diffusioneodistribuzionedeicontenutistessimediantequalunquepiattaforma
tecnologica,supportooretetelematica,senzapreviaautorizzazionescrittada
partedell’autore.
CopyrightNotice
3
ELEDIAResearch Center–AllRightsReserved
Outline
(1)
WhatareEAs?
WhatareEAs?
(2)
HowareEAsdefined?
HowareEAsdefined?
(3)
WhyandwhereusingEAs?
WhyandwhereusingEAs?
(4)
Conclusions
Conclusions
4
ELEDIAResearch Center–AllRightsReserved
What are EAs?
What are EAs?
EAs areMethods fordeterminingtheSolution ofProblems...
… but theProblemsMUST beproperlyreformulatedas
suitable OptimizationProblems
Problem
Problem
Eating
•Goodmeal
•Inalovelyplace
Wheretogo?
Home
McDonald’s
RestaurantinParis
RestaurantinChengdu
Solution
Solution
5
ELEDIAResearch Center–AllRightsReserved
Original Problem
Original Problem
ProblemusuallydefinedbymeansofanObjective and
aSetofConstraints orRequirementstobesatisfied
ProblemStatement
ProblemObjective
ProblemObjective
Eating!
Constraints/Requirements
Constraints/Requirements
•Goodmeal
•Inalovelyplace
6
ELEDIAResearch Center–AllRightsReserved
Optimization Problem
Optimization Problem
CostFunction
AnOptimizationProblem modelstheObjective asaCost
Function tobemin‐maximize
subjecttoaSetofConstraints
orRequirements
ProblemReformulation
Eating!
Constraints/Requirements
Constraints/Requirements
•Goodmeal
•Inalovelyplace
Food‐Quantity
Meal‐Quality>Delicious
Place>HistoricalCity
Constraints
ProblemObjective
ProblemObjective
7
ELEDIAResearch Center–AllRightsReserved
Optimization Problem
Optimization Problem
–
–
An Example
An Example
Griewank function
()
()
()
n
f
N
n
N
n
n
n
ff
100
1
1
2
4000
1
cos1001
−
=
=
∏−−+=Φ
∑
2
=
N
CostFunction:
Constraints:
•Goodmeal
•Inalovelyplace
50
1
≤≤
f
55
2
−≤≤
−
f
2=N
Food‐Quantity
8
ELEDIAResearch Center–AllRightsReserved
Problem Solution
Problem Solution
–
–
An Example
An Example
()
()
()
n
f
N
n
N
n
n
n
ff
100
1
1
2
4000
1
cos1001
−
=
=
∏−−+=Φ
∑
CostFunction
Constraints
•Goodmeal
•Inalovelyplace
50
1
≤≤
f
55
2
−≤≤−
f
Maximization
Maximization
ProblemSolution
ProblemSolution
Minimization
Minimization
ProblemSolution
ProblemSolution
0.0
5.2
2
1
=
=
f
f
SolutionParameters
()
[]
{}
fff
opt
Φ=≡ maxarg
max
()
[]
{}
fff
opt
Φ== minarg
min
9
ELEDIAResearch Center–AllRightsReserved
Optimization
Optimization
Problem
Problem
-
-
Properties
Properties
(
)
0≥Φ
f
(
)
{
}
(
)
{
}
f
f
Φ≡Φ 1minmax
PropertyNo.1
PropertyNo.2
(
)
[]
{}
(
)
[]
{}
fff Φ=Φ= 1minargmaxarg
max
()
[]
{}
(
)
[]
{
}
fff Φ=Φ= 1maxargminarg
min
10
ELEDIAResearch Center–AllRightsReserved
Optimization Problem
Optimization Problem
UnivocallyDetermine thesetofunknownParameters (the
coordinatesofthelocation)satisfyingtheproblemconstraints
bymin‐maximizingthecostfunction
ProblemSolution
Home
McDonald’s
Restaurant
inParis
Restaurant
inChengdu
11
ELEDIAResearch Center–AllRightsReserved
The Nature of Optimization Problems
The Nature of Optimization Problems
- Multiminima Functionals -
ProblemStatement:
Eatinggoodmealinalovelyplace!
McDonald’s
RestaurantinParis
RestaurantinChengdu
Solution
Solution
CostFunction
CostFunction
12
ELEDIAResearch Center–AllRightsReserved
Globalminimum
attractionbasin
Localminimaattractionbasins
LocalMinima
Localminimapointsare
attractorsofthesearchtrajectory
generatedbydeterministic
localsearch
GlobalMinimum
Thesmallestvalue
thatafunctiontakes
onthefunctiondomain
The Nature of Optimization Problems
The Nature of Optimization Problems
- Multiminima Functionals -
13
ELEDIAResearch Center–AllRightsReserved
The Nature of Optimization Problems
The Nature of Optimization Problems
“Theensemble
oftheattractionbasins
formsthefunctionlandscape”
Localminimum
Twolocalminimaareneighborsifandonlyif
theirattractionbasinsareneighbors
[*]R.Battiti,et al.,Reactive SearchandIntelligent Optimization.Operations Research/ComputerScienceInterfaces Series,Vol.45,Springer,November 2008.
14
ELEDIAResearch Center–AllRightsReserved
Local Minima: Benefit or Problem?
Local Minima: Benefit or Problem?
“Thevaluesoftheobjectivefunctionatlocalminima
are(generally)betterthanthevaluesatthestartingpoints”
Inaminimizationproblem,theprobability
oflowvaluesoftheobjectivefunction
islargerforlocalminimathanforapoint
randomlyextractedfromthefunctionlandscape
“Inonesearchforlow‐costsolutions,
samplinglocalminimaismoreeffective
thansamplingrandompoints”
[*]R.Battiti,et al.,Reactive SearchandIntelligent Optimization.Operations Research/ComputerScienceInterfaces Series,Vol.45,Springer,November 2008.
15
ELEDIAResearch Center–AllRightsReserved
“Thememoryaboutthepreviouslyfound
localminima,canbemined toproducebetter
andbetterstartingpoints”
()
fΦ
f
PenaltyTerm
Oncethelocalminimum“a”
isfound,thepenaltyterm
allowstoavoidthesearch
intheattractionbasinof“a”
inthenextiterations.
[*]R.Battiti,et al.,Reactive SearchandIntelligent Optimization.Operations Research/ComputerScienceInterfaces Series,Vol.45,Springer,November 2008.
IteratedSearch
Local Minima:
Local Minima:
Tabu
Tabu
LEARNING from
theprevioushistory
16
ELEDIAResearch Center–AllRightsReserved
Optimization Meaning/Needs
Optimization Meaning/Needs
“DefinitionoftheBoundaryLandscape”
Analyticalframework A‐prioriinformation
Eating!
“Icaneatathome!”
TheImportanceofKnowing
ProblemBounds
“OptimizationasInformationExploitation”
17
ELEDIAResearch Center–AllRightsReserved
Iteration1
Optimization as Information Exploitation
Optimization as Information Exploitation
Openthedoor
TheImportanceofKnowing
InformationAcquisition
Main
Main
Entrance
Entrance
18
ELEDIAResearch Center–AllRightsReserved
Iteration2
Icannoteathere
Not there!
InformationAcquisition‐ Tabu
TheImportanceofKnowing
TvRoom
TvRoom
Optimization as Information Exploitation
Optimization as Information Exploitation
19
ELEDIAResearch Center–AllRightsReserved
Iteration3
Kitchen
Kitchen
TheImportanceofKnowing
Thereisfood!
InformationAcquisition
Optimization as Information Exploitation
Optimization as Information Exploitation
20
ELEDIAResearch Center–AllRightsReserved
“Theflyisn’tintelligent.
Attractedbythelightbulb,
itisrepeatedlyburnt”
TheImportanceofLearning
MemorySetup
Optimization as Information Exploitation
Optimization as Information Exploitation
剩余38页未读,继续阅读
资源评论
Dreama_CS
- 粉丝: 3
- 资源: 11
上传资源 快速赚钱
- 我的内容管理 展开
- 我的资源 快来上传第一个资源
- 我的收益 登录查看自己的收益
- 我的积分 登录查看自己的积分
- 我的C币 登录后查看C币余额
- 我的收藏
- 我的下载
- 下载帮助
最新资源
- 最通俗易懂的 MongoDB 非结构化文档存储数据库教程.zip
- 以mongodb为数据库的订单物流小项目.zip
- 腾讯云-mongodb数据库, 项目部署.zip
- 腾讯 APIJSON 的 MongoDB 数据库插件.zip
- 理解非关系型数据库和关系型数据库的区别.zip
- 操作简单的Mongodb网页web管理工具,基于Spring Boot2.0支持mongodb集群.zip
- tms-mongodb-web,提供访问mongodb数据的REST API和可灵活扩展的mongodb web 客户端.zip
- SpringBoot整合mongodb学习MongoTemplate和MongoRepository两种方式CRUD使用.zip
- SpringBoot整合MongoDB实现对数据库的CRUD小demo.zip
- Python操作MongoDB数据库的基本一些操作 .zip
资源上传下载、课程学习等过程中有任何疑问或建议,欢迎提出宝贵意见哦~我们会及时处理!
点击此处反馈
安全验证
文档复制为VIP权益,开通VIP直接复制
信息提交成功