ebook img

Linear Programming Computation PDF

739 Pages·2023·6.82 MB·English
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview Linear Programming Computation

Ping-Qi PAN Linear Programming Computation Second Edition Linear Programming Computation Ping-Qi PAN Linear Programming Computation Second Edition Ping-QiPAN DepartmentofMathematics SoutheastUniversity Nanjing,China ISBN978-981-19-0146-1 ISBN978-981-19-0147-8 (eBook) https://doi.org/10.1007/978-981-19-0147-8 ©TheEditor(s)(ifapplicable)andTheAuthor(s),underexclusivelicensetoSpringerNatureSingapore PteLtd.2023 Thisworkissubjecttocopyright.AllrightsaresolelyandexclusivelylicensedbythePublisher,whether thewhole orpart ofthematerial isconcerned, specifically therights oftranslation, reprinting, reuse ofillustrations, recitation, broadcasting, reproductiononmicrofilmsorinanyotherphysicalway,and transmissionorinformationstorageandretrieval,electronicadaptation,computersoftware,orbysimilar ordissimilarmethodologynowknownorhereafterdeveloped. Theuseofgeneraldescriptivenames,registerednames,trademarks,servicemarks,etc.inthispublication doesnotimply,evenintheabsenceofaspecificstatement,thatsuchnamesareexemptfromtherelevant protectivelawsandregulationsandthereforefreeforgeneraluse. Thepublisher,theauthorsandtheeditorsaresafetoassumethattheadviceandinformationinthisbook arebelievedtobetrueandaccurateatthedateofpublication.Neitherthepublishernortheauthorsor theeditorsgiveawarranty,expressedorimplied,withrespecttothematerialcontainedhereinorforany errorsoromissionsthatmayhavebeenmade.Thepublisherremainsneutralwithregardtojurisdictional claimsinpublishedmapsandinstitutionalaffiliations. ThisSpringerimprintispublishedbytheregisteredcompanySpringerNatureSingaporePteLtd. The registered company address is: 152 Beach Road, #21-01/04 Gateway East, Singapore 189721, Singapore Dedicatedtomy parents,ChaoPan and YinyunYan;mywife, MinghuaJiang;myson, Yunpeng;mygranddaughter,Heidi Preface to First Edition Linear programming(LP) foundedby Dantzig (1948–1951b)mightbe one of the most well-known and widely used mathematical tools in the world. As a branch ofoptimization,itservesasthemostimportantcornerstoneofoperationsresearch, decisionscience,andmanagementscience. This branch emerged when the American mathematician George B. Dantzig created the LP model and the simplex method in 1947. The computer, emerging aroundthesameperiod,propelledthedevelopmentofLPandthesimplexmethod toward practical applications. As a basic branch, LP orchestrated the birth of a number of new branches, such as nonlinear programming, network flow and combinatorial optimization, stochastic programming, integer programming, and complementarytheory,etc.,andactivatedthewholefieldofoperationsresearch. A prominent feature of LP is given by its broad applications. Closely related to LP, many individuals have made pioneering contributions in their respective fields.Inthefieldofeconomics,inparticular,Russian-AmericaneconomistWassily Leontief took the 1973 Nobel Economic Prize for his epoch-makingcontribution toquantitativeanalysisofeconomicactivities.TheacademicianL.V.Kantorovich oftheformerSovietAcademyofScienceandAmericaneconomistProfessorT.C. Koopmanswonthe1975NobelPrizefortheiroptimalallocationtheoryofresources usingLP.ThesameprizewasalsogiventoProfessorsK.Arrow,P.Samuelson,H. Simon,andL.Herwricz,severaldecadeslaterwhentheypaidcloseattentiontoLP atthestartingdaysoftheirprofessionalcareers. Thesimplexmethodhasalsoachievedgreatsuccessinpractice.Asitisknown, itsapplicationstomanyfields,suchaseconomy,commerce,production,scienceand technology, and defense and military affairs, etc, have brought about astonishing economicandsocialbenefits.ItisrecognizedasoneofTheTenAlgorithmsinthe TwentyCentury(IEEE2002;seeCipra(2000). After morethan 70 years since its birth,LP is now a relativelymaturebut still developing branch. Nevertheless, there exist great challenges. The importance of large-scale sparse LP models is nowadaysenhanced further by globalization.The everyday practice calls upon the research community to provide more powerful solutiontoolsjusttokeepupwiththeever-increasingproblemsizes.Therefore,this vii viii PrefacetoFirstEdition book does not only present fundamentalmaterials but also attempts to reflect the stateoftheartofLP.Ithasbeenmylong-lastingbeliefthatresearch,inoperations research/management science, in particular, should be of practical value, at least potentially.Theauthor,therefore,focusesontheories,methods,andimplementation techniquesthatarecloselyrelatedtoLPcomputation. This bookconsistsof two parts. PartI mainlycoversfundamentaland conven- tional materials, such as the geometryof the feasible region,the simplex method, dualityprinciple,anddualsimplexmethod,implementationofthesimplexmethod, sensitivityanalysis,parametricLP,variantsofthesimplexmethod,decomposition method,and interior-pointmethod.Inaddition,integerlinearprogramming(ILP), differingfromLPinnature,isalso discussednotonlybecauseILPmodelscanbe handled by solving a sequence of LP models but also because they are so rich in practiceandformamajorapplicationareaofLPcomputations. Part II presents published or unpublished new results achieved by the author himself,suchaspivotrule,dualpivotrule,simplexphase-1method,dualsimplex phase-I method, reduced simplex method, generalized reduced simplex method, deficient-basismethod,dualdeficient-basismethod,facemethod,dualfacemethod, and pivotal interior-pointmethod. The last chapter contains miscellaneous topics, suchassomespecialformsoftheLPproblem,approachestointerceptingforprimal anddualoptimalsets,practicalpricingschemes,relaxationprinciple,localduality, decompositionprinciple,andILPmethodbasedonthegeneralizedreducedsimplex framework. Tomakethematerialeasiertofollowandunderstand,thealgorithmsinthisbook were formulated and illustrative examples were worked out wherever possible. If thebookisusedasatextbookforupper-levelundergraduate/graduatecourse,(Part I)maybesuitabletobeasbasiccoursematerial. Acknowledgments At this very moment, I deeply cherish the memory of Professor Xuchu He, my formermentorat NanjingUniversity,arousedmy interestin optimization.I honor the Father of linear programming and the simplex method, Professor George. B. Dantzigfortheencouragementgivenduringthe16thInternationalSymposiumon MathematicalProgramming,LausanneEPFL., in August1997.Iam verygrateful toProfessorMichaelSaundersofStanfordUniversityforhisthoughtfulcomments and suggestions given in the past. He selflessly offered the MINOS 5.51 package (the latest version of MINOS at that time) and related materials, from which my workandthisbookbenefitedgreatly.MINOShasbeenthebenchmarkandplatform of our computationalexperiments. I thank Professor R. Tyrrell Rockafellar of the UniversityofWashington,ProfessorThomasF.ColemanofCornellUniversity,and Professor Weide Zheng of Nanjing University for their support and assistance. I wouldalso like to thankthe followingcolleaguesfortheir supportand assistance: Professor James V. Burke of the University of Washington, Professor Liqun PrefacetoFirstEdition ix Qi of Hong Kong Polytechnic University, Professor Lizhi Liao of Hong Kong BaptistUniversity,ProfessorsLeenaSuhlandDr. AchimKobersteinofPaderborn University,ProfessorUweSuhlofFreieUniversity,andDr.PabloGuerrero-Garcia oftheUniversityofMalaga. This book was supported partially by projects 10871043 and 70971136 of the NationalNaturalScienceFoundationofChina. Nanjing,China Ping-QiPan December2012 Preface to Second Edition Sinceitsbirth,linearprogramming(LP)hasachievedgreatsuccessinmanyfields, such as economy, commerce, production, science, and technology, and brought about amazing economic and social benefits. After 70 years of development, LP now becomes a relatively mature branch in OR/MS. Nevertheless, the academic community faces a major challenge of the growing demand, which needs more powerfulandrobusttoolstodealwithlarge-scaleanddifficultproblems. To meet the challenge, this book draws materials from a practical point of view, focusing on theories, methods, and implementation techniques that are closely related to LP computation. Its first edition consists of two parts. Roughly speaking, Part I covers fundamental materials of LP, and Part II includes pub- lished/unpublishednewresultsachievedbytheauthor. Littleattentionwasreceivedviatheauthor’sResearchGateWebpageuntilShi, Zhang,andZhupublishedtheirbookreviewintheEuropeanJournalofOperational Research(June2018).Sincethen,thebookquicklyattractedconsiderableattention fromacademiccommunitiesaroundtheworld.AsofNovember6,2022,thetopten chapterreadsareasfollows: Chapter Reads 1.DualityPrincipleandDualSimplexMethod(PartI) 15419 2.SimplexFeasible-PointxMethod(PartII) 7861 3.IntegerLinearProgramming(ILP)(PartI) 7340 4.ImplementationoftheSimplexMethod(PartI) 3284 5.DualSimplexPhase-IMethod(PartII) 2613 6.SimplexMethod(PartI) 2576 7.VariantsoftheSimplexMethod(PartI) 2328 8.GeometryoftheFeasibleRegion(PartI) 1951 9.PivotRule(PartII) 1844 10.SimplexPhase-IMethod(PartI) 1506 xi xii PrefacetoSecondEdition The preceding data are consistent with corresponding downloads from the Springerwebsite. Surprisinglyenough,thechapter”DualPrincipleandDualSimplexMethod”not only received the maximum number of reads as a whole but also almost weekly, evenifitisdevotedtosuchaclassicaltopic.Thismightbebecausethediscussions arecomprehensive,comparedwithpopularliterature. Proposedforthefirsttime,theFeasible-PointSimplexMethod,whichappeared as the final section of the chapter ”Pivotal Interior-Point Method,” received the second-highest reads as a whole. It achieved this in a short period of 10 weeks. Indeed,supportedbythesolidcomputationalresults,themethoditselfmightexceed allexpectations. ThefinalchapterofPartI,”IntegerLinearProgramming(ILP),”usuallyreceived thesecond-highestreadsweekly.Thishappened,Iguess,becausemanyresearchers are interested in the newly proposed controlled-branch method and controlled- cuttingmethod,whichhavepotentialapplicationsfordevelopingnewILPsolvers. The author’s main contributions seem to have not been fully valued by now, includingmostpivotrules,reducedandD-reducedsimplexmethods,deficient-basis anddualdeficient-basismethods,faceanddualfacemethods,andthedecomposi- tionprinciple,includedinPartII.Thisissomewhatsurprisingsincethesemethods are supported by solid computational results, except for the newly introduced reduced and D-reducedsimplex methods, and the decompositionprinciple. In the author’s view, the reduced simplex method is particularly noteworthy. It is the firstsimplexmethodthatsearchesalongtheobjective-edge.Incontrasttoexisting simplex methods, it first determines pivot row and then pivot column, and does sowithoutanyselectionessentially.Asadualversionofit,theD-reducedsimplex methodsharessimilarattractivefeatures,throughdeterminingpivotcolumnfirstand pivotrowlater.Asforthedecompositionprinciple,itallowsforsolvingarbitrarily large-scale(evendense)LPproblems,inessence,givingaglimmeroflightonsome othertypesofseparablelarge-scaleproblems.Timewilllikelyclarifythevalueof thesemethodsvsotherLPsolvers. Wearealsooptimisticaboutthefaceanddualfacemethods.Withoutexploiting sparsity, the original methods were implemented using orthogonaltransformation with favorable computational results reported. In the second edition, we added twochaptersdevotedtonewmethodswithLUfactorizationforsparsecomputing. Indeed, this is a very natural development which the author cannot help himself frombreakinginto. Somechangesonchaptersandsectionsweremadewithcorrectionsandimprove- ments,andthewholesecondeditionwasorganizedintotwoparts.Roughlyspeak- ing,PartI(Foundations)containsPartIofthefirstedition,andPartII(Advances) includesPartII.TheSimplexFeasible-PointAlgorithmwasimproved,andremoved from the chapter ”Pivotal Interior-Point Method” to form an independentchapter with its new title “Simplex Interior-Point Method,” since it actually represents a newclassofinterior-pointalgorithms,whichcanbetransformedfromthetraditional simplexalgorithms.Thetitleoftheoriginalchapterwaschangedto“FacialInterior- Point Method,” since the remaining algorithms represent another new class of

See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.