进化算法概述进化算法概述_进化算法资源-CSDN文库

遗传算法

全局优化

需积分: 5 81 浏览量 2018-08-02 20:16:34 上传评论收藏 1.8MB PDF 举报

资源推荐

资源详情

资源评论

ELEDIAResearchCenter

ELEDIA@UniTN ‐ UniversityofTrento

ViaSommarive 9,I‐38123Trento,Italy

E‐mail:andrea.massa@unitn.it

Web:www.eledia.org/eledia‐unitn

Laboratoire desSignauxetSystèmes

UMR8506(CNRS‐CENTRALESUPELEC‐UNIV.PARISSUD)

rueJoliot‐Curie3,91192Gif‐sur‐Yvette,France

E‐mail:andrea.massa@l2s.centralesupelec.fr

Web:www.eledia.org/eledia‐l2s

Prof.AndreaMASSA

OptimizationTheory,Techniques,

andMATLABsimulations

Global(Evolutionary) OptimizationMethods

UESTC“2018 InternationalSummerSchool”

July16‐20, 2018 – Chengdu,China

ELEDIAResearch Center–AllRightsReserved

Allrightstothecontentofthisdocumentarereserved.Anyuse,inwholeor

inpart,ofthecontentsincludedinthisdocument,includingthestorage,

reproduction,editing,disseminationordistributionoftheircontentthroughany

technologyplatform,support,orcomputernetworkisforbiddenwithoutthe

priorwrittenpermissionfrom

theauthor.

Tuttiidirittirelativiaicontenutidelpresentedocumentosonoriservati.È

vietatoqualsiasiutilizzo,totaleoparziale,deicontenutiinseritinelpresente

documento,iviinclusalamemorizzazione,riproduzione,rielaborazione,

diffusioneodistribuzionedeicontenutistessimediantequalunquepiattaforma

tecnologica,supportooretetelematica,senzapreviaautorizzazionescrittada

partedell’autore.

ELEDIAResearch Center–AllRightsReserved

Outline

(1)

WhatareEAs?

(2)

HowareEAsdefined?

(3)

WhyandwhereusingEAs?

(4)

Conclusions

ELEDIAResearch Center–AllRightsReserved

What are EAs?

EAs areMethods fordeterminingtheSolution ofProblems...

… but theProblemsMUST beproperlyreformulatedas

suitable OptimizationProblems

Problem

Eating

•Goodmeal

•Inalovelyplace

Wheretogo?

Home

McDonald’s

RestaurantinParis

RestaurantinChengdu

Solution

ELEDIAResearch Center–AllRightsReserved

The Nature of Optimization Problems

“Theensemble

oftheattractionbasins

formsthefunctionlandscape”

Localminimum

Twolocalminimaareneighborsifandonlyif

theirattractionbasinsareneighbors

[*]R.Battiti,et al.,Reactive SearchandIntelligent Optimization.Operations Research/ComputerScienceInterfaces Series,Vol.45,Springer,November 2008.

ELEDIAResearch Center–AllRightsReserved

Local Minima: Benefit or Problem?

“Thevaluesoftheobjectivefunctionatlocalminima

are(generally)betterthanthevaluesatthestartingpoints”

Inaminimizationproblem,theprobability

oflowvaluesoftheobjectivefunction

islargerforlocalminimathanforapoint

randomlyextractedfromthefunctionlandscape

“Inonesearchforlow‐costsolutions,

samplinglocalminimaismoreeffective

thansamplingrandompoints”

[*]R.Battiti,et al.,Reactive SearchandIntelligent Optimization.Operations Research/ComputerScienceInterfaces Series,Vol.45,Springer,November 2008.

ELEDIAResearch Center–AllRightsReserved

“Thememoryaboutthepreviouslyfound

localminima,canbemined toproducebetter

andbetterstartingpoints”

()

fΦ

PenaltyTerm

Oncethelocalminimum“a”

isfound,thepenaltyterm

allowstoavoidthesearch

intheattractionbasinof“a”

inthenextiterations.

[*]R.Battiti,et al.,Reactive SearchandIntelligent Optimization.Operations Research/ComputerScienceInterfaces Series,Vol.45,Springer,November 2008.

IteratedSearch

Local Minima:

Tabu

LEARNING from

theprevioushistory