Upload
others
View
2
Download
0
Embed Size (px)
Citation preview
— courseoverview—18.06:LinearAlgebra
Prof.StevenG.Johnson,MITAppliedMathFall2017
http://web.mit.edu/18.06
Textbook:Strang,IntroductiontoLinearAlgebra,5th edition+supplementarynotes
Helpwanted:
arrive10minutesearlyandgetpaid$10toerasetheboards
(Youcanforeverafterputblackboardmonitor,onyourresumé
…shout-outtoTerryPratchettfans…)
AdministrativeDetails
LecturesMWF11,54-100Tuesdayrecitations— useStellartoswitchsections
Weeklypsets,dueWednesday11aminyourrecitationbox.— noextensions ormakeup,butlowestpset scorewillbedropped— pset 1tobepostedshortly
Grading:homework15%,3exams45%(9/25,10/30,&11/27in50-340),&finalexam40%
Collaborationpolicy:talktoanyoneyouwant,readanythingyouwant,but:•Makeaneffortonaproblembeforecollaborating.•Writeupyoursolutions independently (from“blanksheetofpaper”).•Listyourcollaboratorsandexternal sources (notcoursematerials).
SyllabusandCalendar• SignificantoverlapwithStrang’s OCWvideo lectures: theseareauseful
supplement butnotareplacement forattending lecture.Likelytopics:• Exam1:Monday9/25.Elimination, LUfactorization,nullspaces andother
subspaces, basesanddimensions, vectorspaces,complexity. (Book:1–3.5,11.1)• Exam2:Monday10/30.Orthogonality,projections, least-squares, QR,Gram–
Schmidt,orthogonal functions (Book:1–4,10.5).• Exam3:Monday11/27. Eigenvectors,determinants, similarmatices,Markov
matrices,ODEs,symmetricmatrices, definitematrices,matrices fromgraphsandengineering. (Book:1–7,10.1–3.)
• Othertopics:defectivematrices, SVDandprincipal-components analysis, sparsematrices anditerativemethods,complexmatrices, symmetric linearoperatorsonfunctions.
• Finalexam:alloftheabove.
Whatis18.06about?2x1 + 4x2 − 2x3 = 24x1 + 9x2 −3x3 = 8−2x1 −3x2 + 7x3 =10
Highschool:3“linear”equations(only± and× constants)in3unknowns
Method:eliminateunknownsoneatatime.
Equivalentmatrix problem
Ax=b
Axisa“linearoperation:”A(x+y)=Ax+AyA(3x)=3Ax
2 4 −24 9 −3−2 −3 7
⎛
⎝
⎜⎜⎜
⎞
⎠
⎟⎟⎟
x1x2x3
⎛
⎝
⎜⎜⎜⎜
⎞
⎠
⎟⎟⎟⎟=
2810
⎛
⎝
⎜⎜⎜
⎞
⎠
⎟⎟⎟
A x b
take“dotproducts”ofrows× columns
matrixofcoefficients
vectorofunknowns
vectorofright-handsides
Whatis18.06about?
Linearsystemofequations,in matrix form
Ax=b
2 4 −24 9 −3−2 −3 7
⎛
⎝
⎜⎜⎜
⎞
⎠
⎟⎟⎟
x1x2x3
⎛
⎝
⎜⎜⎜⎜
⎞
⎠
⎟⎟⎟⎟=
2810
⎛
⎝
⎜⎜⎜
⎞
⎠
⎟⎟⎟
A x b
Willwelearnfastermethods tosolvethis? No.(Except ifAisspecial.) Thestandard“Gaussianelimination” (and“LUfactorization”)matrixmethodsare justaslightlymoreorganized versionofthehigh-schoolalgebraelimination technique.
Willwegetbetteratdoingthesecalculations byhand?Maybe,butwhocares?Nowadays,allimportantmatrixcalculations aredonebycomputers.
Willwelearnmoreaboutthecomputeralgorithms? Alittle. Butmostlythetechniquesfor“serious”numerical linear-algebra aretopicsforanothercourse(e.g.18.335).
Howdowethink aboutlinearsystems?(imaginesomeonegivesyoua106×106 matrix)
• Alltheformulasfor2×2and3×3matriceswouldfitononepieceofpaper.Theyaren’tthereasonwhylinearalgebrais important(asaclass orafieldofstudy).
• Largeproblemsaresolvedbycomputers,butmustbeunderstoodbyhumanbeings. (Andweneedtogivecomputerstherighttasks!)
• Understandnon-squareproblems:#equations >#unknownsorviceversa
• Breakupmatrices intosimpler pieces– Factorizematricesintoproductsofsimplermatrices:A=LU(triangular:Gauss),A=QR
(orthogonal/triangular),A=XΛX–1 (diagonal:eigenvecs/vals),A=UΣV*(orthogonal/diagonal:SVD)
– Submatrices (matricesofmatrices).
• Breakupvectorsintosimplerpieces: subspaces andbasischoices.
• Algebraicmanipulations toturnharder/unfamiliar problems(e.g.minimization ordifferentialequations) intosimpler/familiar ones:algebraonwholematrices atonce
Don’texpectalotof“turnthecrank”problemsonpsets orexamsoftheform
“solvethissystemofequations.”
…wewillturnitupside-down,giveyoutheanswerandaskthequestion,askaboutpropertiesofthesolutionfrompartial
information,…thegeneralgoalistorequireyoutounderstandthecrankratherthanjustturnit.
(Examsmightbeabitharderthaninprevious18.06semesters,butgradecutoffswillbe
adjustedaccordingly.)
Wheredobigmatricescomefrom?
Lotsofexamplesinmanyfields,buthereareacouplethatarerelativelyeasytounderstand…
Engineering&ScientificModeling[18.303,18.330,6.336,6.339,…]
http://gmsh.info/ Unknownfunctions(fluidflow,mechanicalstress,electromagnetic fields,…)approximated byvaluesonadiscretemesh/grid
e.g.100x100x100grid=106 unknowns!
http://www.electronics-eetimes.com/ comsol.com
Dataanalysis
Imageprocessing:imagesarejustmatrices ofnumbers
(red/green/blue intensity)
[Mathw
orks]
[wikipedia ]
Regression:(curvefitting)
Google“pagerank”problem(alsoforgenenetworksetc.)
Determine the“mostimportant”webpages justfromhowtheylink.
matrix=(#webpages)× (#webpages)(entry=1iftheylink,0otherwise)
Notjustmatricesofnumbers• Therearelotsofsurprisingandimportantgeneralizations oftheideasinlinearalgebra.
• Insteadofvectors withafinitenumberofunknowns,similarideasapplytofunctionswithaninfinitenumberofunknowns.
• Insteadofmatrices multiplyingvectors,wecanthinkaboutlinearoperatorsonfunctions
“A” “x” “b”linear
operator∇2
unknownfunctionu(x,y,z)
right-handsidef(x,y,z)
∇2u = fPoisson’sequation(e.g.18.303)
18.06 vs.18.700
“applied”vs.“pure”math
fewproofsvs.formalproofsexpected(deducepatterns (definitions tofromexamples, lemmas totheorems
informalarguments) …traininginproofwriting)
moreapplications vs.moretheoremsmoreconcrete vs.moreabstract
somecomputersvs.onlypencil-and-paper
Computersoftware
[image:ViralShah]
Lotsofchoices:
julialang.org
Thissemester: arelativelynewlanguagethatscalesbetter torealproblems.
Noprogrammingrequiredfor18.06,justa“glorified calculator”toturnthecrank.
Use itonline: loginatjuliabox.comsee“Julia” linkonStellar
Optionaltutorial:Monday5pm32-xxx