人工智能.docx
- 文档编号:7097217
- 上传时间:2023-01-17
- 格式:DOCX
- 页数:12
- 大小:27.27KB
人工智能.docx
《人工智能.docx》由会员分享,可在线阅读,更多相关《人工智能.docx(12页珍藏版)》请在冰豆网上搜索。
人工智能
IntroductiontoExpertSystems
1.1Introduction
Thecomputerhasplayedalargepartinadvancingtechnologyinnovationinmanypartofscienceandindustry.Insomeareashowever,thosewhichtendtoconsistofproblemsrequiringsignificantinputfromhumanexpertfortheirsolution,ithasbeenofmorelimitedsuccess.Thedevelopmentofcomputersystemswhichcanmanipulateexperienceandknowledgeisleadingtonewcomputerapplicationinbranchesofminingthathave,sofar,failedtobenefitsubstantiallyfromcomputertechnology.These‘ExpertSystems’,whichoriginatedinmajorresearchproblemsaimedatdeveloping‘ArtificialIntelligence’,arenolongerresearchtoolsbutarecommerciallyavailableproblem-solvingaidswhichcanbeutilizedinawiderangeofapplication.
1.2ExpertiseandKnowledge
Changesintechnologyareoccurringatanacceleratingrate.Inthepast,timeisavailableforengineersandmanagementtobecomeaccustomedtonewequipmentandtechniques.Today,eventhemostaccomplishedexpertsinafieldoftenhavedifficultykeepingpacewiththefinedetailsofadvances.Tworesearches,expertiseandknowledge,canbeidentifiedasbeingespeciallycriticalinarapidlychangingtechnologically-basedworld.
1.2.1Expertise
Anexpertisapersonwhohashadsufficientexposuretoasubject,suchthathehasdevelopedexpertiseinthatsubjectthisexpertise,orknow-how,utilizedbytheexpertduringthecourseofhisduties.Anexponential-typegrowthintechnology,togetherwithprolongedtrainingexperiencegatheringperiods(10to15yearsformostexperts)isleadingtothe‘ExpertiseVacuum’.Trueexpertsareinshortsupplyandcannotbeproducedquicklyenoughtokeeppacewiththetechnologyexplosion.
1.2.2Knowledge
Assumingthatanexpertknowshowtocarryoutaparticulartask,itisimportantthathehassufficientinformation,orknowledge,aboutaparticularproblemwhenteyingtomakedecisions.Managers,forexample,usuallywishformoreinformationbeforemakingimportantdecisions.Managersrequireup-to-dateknowledge,notsimplydate,aboutthecriticalcomponentsofanoperation,iftheyaretofunctioneffectively.
Itisanareasrequiringexpertiseandknowledgethatconventionalcomputerprogramsareweakandwhereexpertsystemsofferpotentialbenefits.
1.3DisadvantagesofConventionalPrograms
Expertsystemsdifferfromconventionalprogramsinanumberoffundamentalways.Themaindisadvantagesofconventionalprogramsinclude;
a)Traditionalprogramsworkusinganalgorithmic
approachwhichwill,assumingasuitablealgorithmhasbeenchosen,providethecorrectsolutiontoaproblemifgivenALLthecorrectdate.Theyairvirtuallyuselesswheninformationismissingandusuallyunabletoprovidepart-answertoproblems.
b)Thestructureoftheprogramisfixedanddeterminedbytheprogrammer.Conventionalprogramshavetypicallylongdevelopmenttimeswhichresultsinhighcostandslowimplementation.Theprogrammersisalsoresponsibleforprogrammaintenance.Modificationtoaconventionalprogram,ifpossible,isoftentimeconsumingandcostly.
c)Conventionalprogrammingisgoodwithnumericdatebathandlessymbolicinformationpoorly.Realproblemrarelyinvolveonlynumericdate.
1.4ExpertSystems
ExpertSystems,orKnowledge-basedExpertSystems(KBES)astheyareoftentermed,havebeenformallydefinedasintelligentcomputerprogramsthatuseknowledgeandinferenceprocedurestosolveproblemsthataredifficultenoughtorequiresignificanthumanexpertisefortheirsolution.Theyareoneofthefirstusablespin-offsfromtheinternationalresearchprogramsinvestigationthescienceof‘ArtificialIntelligence’whichinvolvessuchareasasnaturallanguageprocessing,imageprocessingandrobotics.Figure1.1showswhereexpertsystemsfitintotheoverallschemeofcomputerscience.
Figure1.1ExpertSystemsinComputerScience
Numerousdefinitionofknowledge-basedsystemhavebeengivenincluding;
(Fiegembaum1983)
Aknowledge-based(expert)system(KEBS)isacomputerprogramsataskknormallydonebyanexpertorconsultantandwhich,insodoingusescaptured,heuristicknowledge.
(Gaschnig1981)
Exertsystemareinteractivecomputerprogramsincorporatingjudgement,rulesofthumb,intuitionandotherexpertisetoprovideknowledgeableadviceaboutavarietyoftasks.
(Brachman1983)
Anexpertsystemisonethathasexpertrulesandavoidsblindsearchbymanipulatingsymbols,graspsfundamentaldomainprinciples,andhascompleteweakerreasoningmethodstofallbackonwhenexpertrulesfailandtouseinproducingexplanations.Itdealswithdifficultproblemsinacomplexdomain,cantakeaproblemdescriptioninlaytermsandconvertitintoaninternalrepresentationappropriateforprocessingwithitsexpertrules,andcanreasonaboutitsitsownknowledge(orlackthereof),especiallytoreconstructinferencepathsrationallyforexplanationandself-justification.
Expertsystemsareexpandingtheapplicationsforcomputersintonewareasofproblemsolving.Theseareasarethosewhichinvolveformorethanthenumericaldateprocessingofearliercomputingsystems.Whilsttheconventionalcomputerprogrammingapproachisstillvalidinmanytypesofproblemandwillcontinuetobesowellintothefuture,itisintheapplicationofcomputerstonewproblemfieldsthattheyarebeingsurpassedbyexpertsystems.
1.4.1DifferencebetweenConventionalProcessingandKEBSProcessing
Conventionalprocessing
1)Ifwethinkofacomputerprogramasacollectionofrulesthatdescribeactionstobetaken,thenthetraditionalalgorithmicprogrammaybeviewedasonewherethesequenceoffiring(ortestingandapplicationof)therulesisdeterminedinadvanceandwhereeachrulepremiseorconditionleadstooneandonlyoneaction.
2)Furthertheprocessingisdoneusingsymbolstorepresentnumbers,arithmeticpropertiesandmathematicaloperations.
3)Utilisesproceduralprogrammingwheretellsthecomputerwhatmustbedonewiththedatesubsequentlyentered.Thus,theproceduresarearepresentationofhowwewantsomethingdone.
KBESProcessing
1)Incontrast,inaknowledge-basedsystem,thesequenceofrulefiringisdeterminedbyaninferenceenginethatiscontainedwithintheproblem,andtheconditionsrequiredtofireanyrulesmayleadtomultipleactionsortonoactionatall.Thecollectionsofrulesinsuchasystemwillusuallyincorporateheuristicsorrulesofthumbthatareaccumulatedbyanexpertoveryearsofproblemsolving.
2)Thesymbolsinaknowledge-basedsystemareusedtorepresentvirtuallyanykindofobject,includingpeople,materials,conceptsetc.
3)Utilizesdeclarativeknowledgewillbeused.
1.4.2FeaturesofExpertsystems
Expertsystemsaredesignedtooperatelikehumanexperts.Thefeaturesthatarecommontomostexpertsystemsinclude;
a)Theyarebasedonheuristicapproachestoproblemsolving.Facts,rules-ofthumbandrelationshipsarethecomponentsofthesesystemsratherthannumericdateandalgorithms.Theycan,infact,manipulatesymbolicinformationandrulesmoreeasilythannumericdateandalgorithms.
b)Theyconsultwiththeusertodefineandsolveaproblem.Thenatureoftheconsultationvariesfromproblemtoproblemanddependsontheinformationavailableandtheresponseoftheusertoquestioningbythesystem.
c)Theycanexplaintheirreasoningbehindanyconclusiontheyhavereached,ifasked.Thereasoningatanystageofaconsultationcanalsobeobtainedandtheycanusuallyexplainwhytheyareaskingaparticularquestion.
d)Theycanfunctionusefullywithonlypartialinputdateandevenwhennotfullydeveloped.Theycanalsomanipulateandhandlethesortofuncertaininformationthatisoftenavailablewhentryingtosolveproblems.
e)Mostsystemsattempttocarryoutconsultationsandprovideexplanationsorconclusionsina‘naturallanguage’format.
f)Theycanoftenbemodifiedeasilybynon-programmingpersonnel,increasingtheflexibilityofthesystemandreducingdependenceontrainedprogrammingstaff.
1.4.3AdvantagesofExpertSystems
Figure1.2illustratessomeoftheadvantagesofexpertsystemswhichinclude;
a)Theyarepermanent,thatistosaythatoncedevelopedtheyprovideastoreofexpertisethatcanberetainedbytheorganizationevenifthehumanexpertleavestheorganization;
b)Theyareconsistent,inthatgiventhesamesetofconditionsdescribingaproblemtheywillalwaysprovidethesamesolution.Humanexpertdonotalwaysexhibitthistrait.
c)Theyareaffordable,Oncedevelopedtheyarecheapandeasytodistributethusreducingthecostofexpertadvicetotheorganization.
d)Theyareexpandabletoallowabankofexpertknowledgetobebuiltwhichmayfarexceedthecapabilitiesofasingleexpert.
e)Theyarealwaysavailable.Incriticalsituationstheabilitytoaccessexpertknowledgewithouttheneedtocalloutthehumanexpertisvital.Inadditionexpertsystemstendnottorequireholiday!
f)Theyareeasilyaccessible.Ifdevelopedproperlyandpersonnelaretrainedproperlythereshouldbefewoftheinhibitionsconcerningthefearofaskingsimplequestionsofthehumanexpert.
g)Theyarespeedy.Thebenefitsofcomputertechnologymeanthatanswerscanbeprovidedinseconds.
1.5PotentialApplicationsAreas
1.5.1ProblemCategories
Amarel(1978)defined2maincategoriesofproblems(figure1.3)
DerivationorClassification
FormationorSynthesis
Figure1.3CategoriesofProblem
Derivation
Solutionsarederivedfromgivenfactsanddate.Handingofunreliableandtime-varyingdate.
Diagnosis-Identifythereasonsforfailureormalfunction.
Interpretation-Discernapatternamongasetofdate.
Monit
- 配套讲稿:
如PPT文件的首页显示word图标,表示该PPT已包含配套word讲稿。双击word图标可打开word文档。
- 特殊限制:
部分文档作品中含有的国旗、国徽等图片,仅作为作品整体效果示例展示,禁止商用。设计者仅对作品中独创性部分享有著作权。
- 关 键 词:
- 人工智能