Sunteți pe pagina 1din 17

EECS203finalexamstudyguide

winter 2012, University of Michigan

byEvanHahn+ScottGodbold+BradHekman+AlexIhlenburg+RyanYezman+Matt
Schulte+LuluTang+AndyModell+KevinByung+DavidBrownman+OttoSipe+Yaoyun
Shi+ThomasLovett+SeanHacker+MikeCuskley&DanielNees
(addyournameshereifyouhelped!)

Feelfreetoeditifyouthinkyouhavesomethingtoadd,butpleasedonotruinthis
voluptuous document.

Thisstudyguideisfreelicenseinthepublicdomain.Itsaccuracyisnotguaranteed.

EXAMINFO
Youcanbring3singlesidedsheets
"Inaddition,inyoursolutionsyouareallowedtoquotewithoutproofanyresultsproved
inthelectures,discussionsections,homework,thesampleexam,andthetextbook."

PROPOSITIONSANDOPERATIONS

Propositioneithertrueorfalse,butnotboth
p=negationofp
pq=and(rememberthatisliketheAin"AND")
pq=or
pq=ExclusiveorXORonlyonecanbetrue
Tautologyalwaystrue
Contradictionalwaysfalse
Distributivelaws
p(qr)(pvq)(pvr)
p(qvr)(pq)v(pr)

DeMorgan'sLaws
(pq)pvq
(pq)pq
1

CONDITIONALS
pq:ifp,thenq
TF=Feverythingelseistrue

Englishwaystoexpressconditionals

"ifp,thenq"
"ifp,q"
"pissufficientforq"
"qifp"
"qwhenp"
"anecessaryconditionforpisq"
"qunlessp"
"pimpliesq"
"ponlyifq"
"asufficientconditionforqisp"
"qwheneverp"
"qisnecessaryforp"
"qfollowsfromp"

Logicalequivalences

pqpvq
pqqp
pvqpq
pq(pq)
(pq)pq
(pq)(pr)p(qr)
(pr)(qr)(pvq)r
(pq)v(pr)p(qvr)
(pr)v(qr)(pq)r

Biconditionallogicalequivalences

pq(pq)(qp)
pqpq
pq(pq)v(pq)
(pq)pq

RULESOFINFERENCE
Forpropositionallogic
[p(pq)]q
[q(pq)]p

(modusponens)
(modustollens)
2

[(pq)(qr)](pr)
[(pvq)p]q
p(pvq)
(pq)p
[(p)(q)](pq)
[(pvq)(pvr)](qvr)

(hypotheticalsyllogism)
(disjunctivesyllogism)
(addition)
(simplification)
(conjunction)
(resolution)

Forquantifiers
xP(x)P(c)(universalinstantiation.Example:"Allwomenarewise",therefore"Lisa
iswise.")
P(c)forarbitrarycxP(x)(universalgeneralization)
xP(x)P(c)forsomec(existentialinstantiation)
P(c)xP(x)forsomec(existentialgeneralization)

QUANTIFIERS
xP(x)=P(x)istrueforeveryxonthedomain
xP(x)=thereisanxonthedomainforwhichP(x)istrue
!xP(x)=thereisONLYONExonthedomainforwhichP(x)istrue

DeMorgan'sLaws
xP(x)xP(x)
xP(x)xP(x)

Nestedquantifiers
xxQ(x)x(xQ(x))

Quantifyingtwovariables

AhelpfulPowerPoint
xyP(x,y)yxP(x,y)=truewhenP(x,y)istrueforeverypairofxandy
xyP(x,y)=truewhen,foreveryx,thereisayforwhichP(x,y)istrue
xyP(x,y)=truewhenthereisanxforwhichP(x,y)istrueforeveryy
xyP(x,y)yxP(x,y)=truewhenthere'sapairx,yforwhichP(x,y)istrue

PROOFS
Proofbyexhaustion:tryeverything(onlyworkswithafinitenumberofpossibilities)
Proofbycases:youcanbreakthingsintogroupsandshowthatcertaingroupswork
Proofbyimplication
AssumePistrue
AssumeP,PthenQ,thereforeQ
HoldsbecauseifPistrueyougetQandifPisfalsethenQis
automaticallyimplied
3

AfteryouproveQitisnolongersafetoassume,Pstillholds
Provethecontrapositive
AssumeNotQshowthatitimpliesNotP
Proving"ifandonlyif"
Proveeachimpliestheother
PimpliesQANDQimpliesP
Constructachainofiffs
ShowPiffRiffSiffQsoPiffQ
Proofbycontradiction
ShowthatifthepremiseisTorFthensomethingFwouldbeT
Example:ShowthatifPisfalse,then0=2since0!=2Pmustbetrue

SETS
Acollectionofofobjectscalledelementsgroupedtogether
Orderingisnotimportant,justwhichelementsarecontained(eg,{x,y}isthesameas
{y,x})
xDshowsxisanelementofD

=theemptyset={}
N=naturalnumbers={0,1,2,...}
Z=integers={...2,1,0,1,2,...}
Z+=positiveintegers={1,2,3,4,...}
Q=rationals
R=reals
R+=positivereals
C=complexnumbers

SetscancontainothersetsdenotedSTshowingSisasubsetofTorallthe
elementsinSareinT
Unioning()twosetscombinesalltheelementsoftwosets(additive)
Intersection()createsasetcontainingallelementscontainedbybothsets
Subtraction()createsasetcontainingallelementsof1stsetexcepttheonesinthe
2ndset.Example:ABcontainsalltheelementsofAexcepttheonesthatareinB.
Thecomplementofasetiseverythingexcepttheelementsinthatset
Asethasacardinalityequaltothenumberofelementsintheset
Example:S={1,2,3}|S|=3
S={{1,2,3},{4,5,6}}|S|=2
S={{},1}|S|=2
S={,1}|S|=2//note:ifiscontainedintheset,thenitcountsasan
element.Ifthesetitselfisempty,thanthecardinalityis0(seebelow).
S=|S|=0
Powerset
Thesetcontainingallsubsetsofaset(Iknow,kindofstrangewording)
ex.BP({1,2})iffB={{},{1},{2},{1,2}}
Cardinalityofapowersetis:|P(S)|=2^|S|
Cartesianproductofsets
4

Givesallpossiblesetsofwhichthefirstelementcomesfromsetoneandthe
secondelementcomesfromset2
Notationfordefiningaset
A={nR|x^33x+1>0}
Setequality
Twosetsareequalifandonlyifeveryelementinonesetisintheotherand
visaversa

INDUCTION
Principle
P(n)isyourpredicate
IfP(0)istrueandP(k)impliesP(k+1)thenP(n)istrueforallnintheset
InductionTemplate
Stateyouareusinginduction

DefineyourpredicateP(n)
ProvethatP(0)istrue(BaseCase)
AssumeP(n)istrue
ShowthatP(n+1)istrue
StrongInduction
P(0)istrueandP(0)throughP(n)implyP(n+1)

FUNCTIONS
AfunctionisamappingfromdomainAtocodomainBsuchthatforanyoneinput
therecanonlybeoneoutput
Denotedf:ABorf(A)={y=f(x)|xA}
Onetoone/injectiveXisonetoonewithYifeverymappingtoYhasexactlyONE
valueinXthathitsit
Onto/surjectiveXisontoYifthereexistvaluesinXsuchthatmappingthemontoY
willhiteveryvalueinsideofY
Bijectiveitgoesbothways:ontoandonetoone
OnetoonecorrespondenceAfunctionisaonetoonecorrespondenceifitis11
andonto
Inversefunctionsafunctionf1(x)themapsbontoaiffthefunctionf(x)mapsfroma
tob.Thisonlyworksiffisontoand11.
CountableainfinitesetwithaonetoonecorrespondencewiththesetofNatural
numbers
CompositefunctionsMappingatocthroughasetoftwofunctionsthatmapatob
andthenbtoc,denotedfog.Otherwisewrittenasf(g(a)).

ALGORITHMS
Algorithmfinitesequenceofpreciseinstructionstoperformacomputation

Searches
Linearsearchstartatthefront,bruteforceuntilyoufindit.Worstcase:O(n)
Binarysearchstartatthemiddleofasortedlist.Ifyouretoohigh,searchtheleft
halfofthelist.Ifyouretoolow,searchtherighthalfofthelist.Worstcase:O(log(n))

Sorts
Bubblesortswitchadjacentvaluesoverandoveragainuntilyoudontswitchno
more.Anotherwaytothinkaboutbubblesortis:findthebiggestelementandputitat
theend,thenlookforthesecondbiggestelementandputitonefromtheendandso
on.
WorstcaseO(n2),bestcaseO(n).
Insertionsortplaceeachelementfromtheunsortedlisttothecorrectplaceinthe
sortedlist.WorstcaseO(n2),bestcaseO(n).

Haltingproblem
Procedurethatcannotbesolvedwithanalgorithm.
H(P,I)definesHasaprocedurewhichtakesintwoarguments.PisaprogramandIis
argumentstoP.HreturnstrueifPwillhaltwithargumentsI,orfalseifPloopsforever
ProofthatH(P,I)cannotbesolvedwithaprocedure:defineK(P)ashaltingifH(P,P)(
aprogramcantakeitselfasanargument)returnsfalse(runforever).K(P)runsfor
everifH(P,P)returnstrue(halting).
IfwepassKitselfK(K).IfKhaltsthenH(K,K)wouldvesaidtheKwouldrunforever
andalsoifKrunsforeverH(K,K)wouldvesaidthatKhalts.Thus,creatinga
contradiction.

SEQUENCES
Asetthatfollowsapattern
Recurrencerelationsexpressesthevalueofanasafunctionofthepreviousterm
(ana).Alwayshasabasecaselikea0=1,a1=1,etc.
Commonpatternsforsequencetermstofollowa+dn,arn,n2,n3,n4,2n,3n,n!,fibonacci
6

SUMMATIONS

Moreoftheseformulascanbefoundusinggoogleifthisdoesntseemcomprehensive
enough.

COSTSOFACOMPUTATION
BigO,big,big
BigOupperboundonthegrowthofthefunction
f(x)=O(g(x))thereforeC*g(x)f(x)
BigOmegalowerboundonthegrowthofafunction
f(x)=(g(x))thereforeC*g(x)f(x)
BigThetaafunctionthatdescribesthesamegrowthorderasfunctionyouare
observing
f(x)=(g(x))thereforeC1f(x)=O(g(x))&C2f(x)=(g(x))
Onlymattersasthefunctionprogressestoinfinity,youareallowedtosayforallx>k
WhendisprovingallbigOorbig,shownosuchwitnessespairs(k,C)exist
Algebrarelations
Addition
f(x)isO(g(x))&e(x)isO(h(x))then(f+e)isO(max(g(x),h(x)))
Scalarmultiplication
Whenf(x)isO(g(x))andthereisaconstantkthenk*f(x)isO(g(x))
Product
f(x)isO(g(x))&e(x)isO(h(x))then(f*e)isO(g*h)
Todetermineafunction'sbigOyoumustfindthesumequivalentvalueand
determiningwhatbindsit

Ifthelimitexistsandisfiniteasxgoestoinfinityoff(x)/g(x)thenf(x)O(g(x))
ifthelimitisinfinite,thenf(x)O(g(x))
RecurrenceRelations
7

f(n)=a*f(n/b)+c
O(n^log_b(a))ifa>1
O(log(n))ifa=1
f(n)=af(n/b)+cn^d[aka...themastertheorem]
O(n^d)ifa<b^d
O(n^d*log(n))ifa=b^d
O(n^(log_b(a)))ifa>b^d

Tractability
Aproblemisconsideredtractableifthealgorithmtosolvetheproblemhas
polynomialorlesscomplexity
Aproblemisconsideredintractableifthealgorithmtosolvetheproblemhad
exponentialorhighercomplexity
Aproblemisconsideredunsolvableifnoalgorithmsolvestheproblem.For
example,theHaltingproblem.

COUNTING
InclusionExclusion
|AB|=|A|+|B||AB|
|ABC|=|A|+|B|+|C||AB||AC||BC|+|ABC|

CartesianProduct
S=S1S2...Sn
|S|=|S1||S2|...|Sn|(whetherdisjointornondisjoint)
Composite:notprime

Pigeonholeprinciple
IfNpigeonsnestink<Nholes,thenatleastoneholewillhavetwoormorepigeons.
Generalizedpigeonholeprinciple
IfNobjectsareplacedintokboxes,thenthereisatleastoneboxcontainingat
leastCEIL(N/k)objects.
Purifiedpigeonholeprinciple(EWD)
Foranonempty,finitebagofnumbers(andbagsallowsrepetitions),the
maximumvalueisatleasttheaveragevalue.

BINOMIALTHEOREM
Definition

n0,(x+y)n= xi y ni (C(n, i))


i=0

Similartothemultinomialtheorem,whichexpandsitouttoanarbitrarymelements
8

COMBINATIONS&PERMUTATIONS
Combinations/choose(akabinomialcoefficient):orderdoesnt
matter
C(n,k)
Thenumberofwaystochoosekobjectsfromnobjectsiforderdoesn'tmatter
Forexample,youwanttopickateamof5from20peopleorderdoesn'tmatter,so
it'sC(20,5)
n!
C(n,k)= k!(nk)!
(andyoucanseeitprettieronWikipedia)
Withrepetitionallowed:C(n,r)=

(n+r1)!
r!(n1)!

Permutations:ordermatters
P(n,k)
Thenumberofwaystochoose(pick)kobjectsfromnobjectsifordermatters
Forexample,ifyouwanttopickthetop10songsfrom200,ordermatters,soit's
P(200,10)
P(n,k)=n!/(nk)!(andyoucanseeitprettieronWikipedia)
Withrepetitionallowed:P(n,k)= nk

PLACINGBALLSINURNS
Placingnballsintomurns(forS(n,m),seebelow):

balls

urns

#balls/urn

#solutions

labeled

labeled

any

mn

labeled

labeled

<=1

P(m,n)

labeled

labeled

>=1

m!S(n,m)

labeled

labeled

kthurnholdsnk
balls

n!/(n1!n2!
nm!)

unlabeled

labeled

any

C(n+m1,m
1)

unlabeled

labeled

<=1

C(m,n)

labeled

unlabeled

>=1

S(n,m)

labeled

unlabeled

unlabeled

unlabeled

any
>=1

S (n, k )
k=1

Pm(n)
9

Stirlingnumbersofthesecondkind
S(n,k)=theamountofwaystosplitndistinguishableobjectsintokindistinguishable
bins
Forexample,ifyouhave4employeesandwanttoputtheminto2offices(andyou
can'tdistinguishtheoffices),that'sS(4,2)=7differentwaystodothis
There'sanuglyformulawhichyoumightwannawritedown

It'susuallybesttouseachart,whichyoumightalsowanttowritedown

PROBABILITY
StartswithadefinedsamplespaceSandaneventEinthatsamplespace
probabilityofEisp(E)=|E|/|S|
0P(E)1
P(notE)=1P(E)
P(E1E2)=P(E1)+P(E2)P(E1E2)
Methodforprobability
Findthesamplespace
Defineeventsofinterest
Determineoutcomeprobabilities
Computeeventprobabilities

BAYES'THEOREM
P(E|F)=theprobabilityofE,givenF

P (E|F ) =

P (F |E)P (E)
P (F |E)(P (E)+P (F |G)(P (G))

GisthecomplementofE
Moregenerally, P (E|F ) =

P (F |E)P (E)
P (F )

.P(F)maybegivenoreasiertodetermine.

IfyouhaveP(F)=,thentheyoucanuseasimplerversion:P(E|F)=P(F|E)/[
P(F|E)+P(F|Ecomplement)]

EXPECTATIONandVARIANCE
Expectedvalue=E(x)
X(s)isarandomvariable...
E(x)=P(si)X(si)
Expectationsarelinear,whetherornotXiareindependent
Usefulinprovingbasicpropositions
E(X1+X2)=E(X1)+E(X2)
E(aX+b)=aE(X)+b
10

TotalExpectations
E(R)=E(R|Ai)X(Ai)
DeviationisX(s)E(x)howeverthisisnotveryusefulandisoftensquared
Variance

Standarddeviation=V(x)

wherep=probability,x=value,mu=E(X)

RELATIONS

Reflexiverelationsyourefriendswithyourself.(A,A)
Symmetricrelationsmarriages.{(A,B),(B,A)}
TransitiveifIvetouchedherhand...{(A,B),(B,C),(C,A)}.(C,A)istransitive.
Antisymmetricif(A,B)and(B,A),thenA=B.(Thatistosay,theonlysymmetry
allowedisreflexiveness.)
Equivalencerelationsaretransitive,symmetric,reflexive
EquivalenceClasssetofelementsrelatedtoa.denoted:[a]
WarshallsTheoremusedtofindapathfromvitovj.ExcellentnotesonWarshalls.
HereisanicelittledemonstrationtocomplementLulus:Warshall

3
Complexity: 2n
PartiallyOrderedSetsorposetarereflexive,antisymmetric,transitive
Givesasetandarelationsharedbytheset,thesecaninclude<,>,=,>=,<=,
|,andmanymore.
WellOrderediftheSfollowsthegivenrelationandeverynonemptysubset
hasaleastelement(noelementsbranchfromit)andifitstotallyordered.
HasseDiagramsshowsaposetthatiswellordereddoesnotshowreflexiveness,or
directionoftheset.
MaximalElementelementsnotunderanyotherelements
MinimalElementelementnotaboveanyelement
GreatestElementeveryelementinthesetislessthanthiselementuniquewhenit
11

exists.
LeastElementeveryelementisgreateruniquealso.
LatticeAposetinwhicheverytwoelementshavealeastupperboundanda
greatestlowerbound.Essentiallyeverypairofnodeshaveaspecificnodethatisa
sharedgreatestorlowestboundthisnodecannotbeonthesamelevelasother
nodesthataregreatestbounds.

Matrixstuff

AdjacencyMatrixHasa1intheijthpositionifvertexiisadjacenttovertexj.
Otherwiseithasa0.(Note:youalwaysreadamatrixinrowcolumnordersotheijth
entryistheithrowandthejthcolumn)
IncidenceMatrixHasa1intheijthpositionifvertexiisconnectedtoedgej.
Otherwise,ithasa0.Thisonerequirestheedgestobelabelled(e1,e2,etc.).
BooleanMatrixAmatrixcontainingonlyzeroesandones,wherezeroesrepresent
falseandonesrepresenttrue.Thisdiffersfromaonezeromatrix.
Composition

GRAPHS
SimplegraphsGraphsthatdonothaveselfloopsorparalleledges.

Vertexapointthatrepresentsaplaceonthegraph(anode)
isolatedvertexofdegree0
pendantiffdegree1
Edgeapathattachingtwovertices
Degreedegreeofvertexvisthenumberofverticesadjacent(itisconnectedto)toit
Degreesequenceanonincreasinglistofvertexdegrees
Specialgraphs
CompletedenotedKnwithnverticesallpairsareconnected.
nvertices
n(n1)/2edges
CycledenotedCnnverticesthatformaclosedpaththatstartsandendson
thesamepoint

12

Thoseareallvalidcycles^
nvertices
nedges
WheeldenotedWnnverticesthatformedbyconnectingasinglevertextoall
theother(n1)vertices

n+1vertices
2nedges
BooleancubedenotedQnwith2nverticesindexedbynbitstringswhere
eachnodediffersitsadjacentnodesby1bit.
2nvertices
n*2n1edges
Regulargraphalldegreesarethesame
example:2regular:allverticeshaveadegreeof2
PetersonGraph
ChromaticNumber:3
3connected
BipartitedenotedKn,mwherethereexistsnoedgeinsidethegraphanditcanbe
separatedintotwosidenandm,isconsideredcompleteifallpairsbetweennandm
areconnected.Itis2colorable.
Chromaticnumbertheminimumnumberofcolorsneededtocoloragraphsuch
thatnotwoadjacentnodeswillhavethesamecolor.
InducedsubgraphPartofagraph.Anedgeinducedsubgraphisalltheedges
listed,aswellastheirendpoints.Avertexinducedsubgraph(thisisthenormalkind)is
alltheverticeslisted,alongwithanyedgesthatgobetweenthem.
MatchingasetofedgesonGsuchthatno2setsshareavertexincommon
Perfectmatchingmatchingwithn/2edges(mustbeanevenamountofverticesfor
aperfectmatching.)
Petersengraph:
edgeconnectivity:3
vertexconnectivity:3
Connectivity
13

Edgetheminimumnumberofedgesremovingwhichwoulddisconnectthe
graph,shown(G)
Vertextheminimumnumberofverticesremovingwhichwoulddisconnect
thegraph,shown(G)
Eulerianpath&cycleapaththatvisitseachedgeonceandonlyonce,itiscalleda
cycleifdoingsomakesacompletecycle
AgraphhasanEulerianpaththatisnotacycleifandonlyifthereareprecisely
twoverticesofodddegree,thispathmuststartononeofthosenodes.
AgraphhasanEuleriancycleifeveryvertexhasanevendegree.
Hamiltonianpath&cycleapaththatvisitseachvertexonlyonce,againisacycleif
doingthisalsomakesacompletecycle.
OrestheoremAgraphGonnverticesisHamiltonianifdeg(u)+deg(v)>=nforall
nonadjacentverticesuandv.
uandvmustnotbeconnected
DiracstheoremAgraphGwithnverticesisHamiltonianif:
Graphissimple
n>=3
degreeofeachvertexisatleastn/2
Stablemarriage:thepreferencerankingmatrixwediscussedatlecture(Allwomen
havethesamelist1,2,..,n,andManislistisi,i+1,...,n,1,2,..,i1).
HallsTheorem:abipartitegraphG=(S,T,E)hasamatchingthatmatchesallvertices
inSiffforallsubsetsWofS,thecardinalityoftheneighborhoodWisgreaterthanor
equaltothecardinalityofW
Proofsketch:keepincreasingthesizeofthematchingthroughaugmenting
path
criticalobjectinproof:augmentingpath
keyfactnumerouno:ifPisanaugmentingpathwithrespecttoamatchingM,
thenMXOPisamatchingand|MXOP|=|M|+1
MXOP=(Pintersect[not]P)union([not]MintersectP)
keystepintheproof:underthemarriagecondition,thereisalwaysan
augmentingpath
intuition:otherwisethesetofverticesreachedbyanalternatingpath
violatesthemarriagetheorem
GaleShapleyTheorem:thedeferredacceptancealgorithm
AcquaintanceshipGraphapathbetweentwopeopleifthereisachainofpeople
linkingthesepeople.Twopeoplewhoareadjacenttoeachotherknoweachother.
AlternatingPathisapathinwhichtheedgesbelongalternativelytothematching
andnottothematching
AugmentingPathsanalternatingpaththatstartsfromandendsonfree(aka.
unmatchedorexposed)vertices.Eachtime,augmentingpathincreasesthetotal
numberofmatchingby1.
Isomorphicchangingthegraphsappearancewithoutchangingthegraphitself.Say
youhaveagraphGandanedgeexistsbetweentwoverticesv1,v2inG.Ifyou
transformthegraph,s.t.v1=xandv2=y,inanewgraphG,thanandanedgeexists
betweenx,yinG.
i)theyhavethesamenumberofedges
ii)theyhavethesamenumberofvertices
iii)theyhavethesamedegreesforallvertices
14

iv)theyhavethesamecircuitsofsamelength
Everyacyclic(nocycles)graphG=(V,E)has|V||E|connectedcomponents
HandshakeTheorem:sumofdegrees=twicethenumberofedges
agraphis2regulariffitsavertexdisjointunionofsimplecycles
forallacyclicgraphs:thenumberofconnectedcomponents+thenumberofedges=
thenumberofvertices
meaningofthepowersoftheadjacencymatrix:theuventryofthekthpoweristhe
numberofuvpathsoflengthk

EXAMPLEPROBLEMS
Q:MontyHallproblem:3doors,onewithacarandtwowithgoats.Youchooseone
withoutopeningitandMontyshowsyoubehindadifferentdoor,onethathasagoat.
Shouldyouchangeyourdoor?
A:Yes.Why?Whenyouchoosefirst,yourchancesare,andsoyourchancesofbeing
wrongare.ButwhenMontyeliminatesadoorthechancethatthedooryoupickedisstill,
andso1=isdistributedacrosstheremainingdoors,inthiscase1door.Soisthe
probabilitythattheremainingdoorhastheprize.

A2:Anotherwaytothinkaboutthis(thatIfoundmorehelpful),isthinkinsteadthatyoure
pickingasinglecarfrom50doors.Youpickdoor42,andMontyopens48otherdoors(allof
whichhavegoats),leaving#1unopened.Now,thechancesofyouhavingpickedthecorrect
doororiginallywere1/50.Theresa49/50chancethatthecarisintheothersetofdoors.The
factthattheyrenowopenonlychangesthefactthattheresa49/50chancethatthecaris
behinddoor#1(asopposedtotheother48possibilities),soyoushouldswitch.

Q:Hatcheckproblem:Youcheckyourhatin,butthehatmanforgottoput
numbersonthehats.Whatistheexpectednumberofpeoplethatgettheirown
hatback?
A:1.Why?Expectationsarelinear.X=X1+X2+X3+X4+X5....+Xn.E(Xi)=1/n.You
canaddupthechancesofeachpersongettingtheirhatbacktogetheexpected
number.Summationof(1/n)fromi=1ton,whichis1.[Forexample3people...1/3+
1/3+1/3=1]

Q:Ifyouthrow2dice,whatistheexpectednumberofthrowsbeforeyougetsnake
eyes.
A:36.Why?Thetotalnumberofways2diecanlandis36,andsnakeeyesonlyappears
in1ofthose36.So1/p=36,p=1/36

Q:TheSAThasameanof500andstandarddeviationof100,whatdoes
Chebyshev'stellusaboutgettingascoregreaterthan700?
A:1/8.Why?P(SAT)>=700=P(SAT>=700or<=300)=V(X)/r^2=*
(100^2)/(200^2)=1/8Theisbecausewewantjustscores>700notlessthan300.100=
thestandarddeviationandthe200comesfromX(s)E(X)whichhereis700500.

15

Q:a)Whatistheprobabilitythattwopeoplechosenatrandomwerebornduringthe
samemonthoftheyear?b)Whatistheprobabilitythatinagroupofnpeoplechosen
atrandomthatthereareatleasttwoborninthesamemonthoftheyear?c)Howmany
peoplechosenatrandomareneededtomaketheprobabilitygreaterthan1/2thatthere
aretwopeopleborninthesameyear.
A:a)1/12.Thereare12monthsintheyear,andbeingbornisanindependentevent.Thatis,
theprobabilitythatyouwereborninacertainmonthdoesntdependontheprobabilitythat
anotherindividualwasborninthatsamemonth.Sincethereisonlyonewayfortwopeopleto
bebornonthesamemonththeprobabilityis1/12.
b)Sameasbirthdayproblem.
c)5.???

Q:a)ProvethatS(n,n1)=C(n,2)forn>=2b)ProvethatS(n,2)=2n11,n>=2
A:a)BythePigeonholePrinciple,aplacementmusthave2ballsinthesameurn,whileby
therequirementthatnournisempty,eachoftheotherurnscontainsprecisely1ball.Thusa
2placementuniquelydetermineapairofballs.Converselyapairofballsuniquelydeterminea
placementinwhichthetwoballsinthepairareinthesameurnandothersintheirotherurn.
ThusthereareC(n,2)placements.
b)EachnonemptypropersubsetAsubsetofnuniquelydeterminesaplacementwithAin
oneurnandB=notAintheotherurn.Foreachplacementprecisely2propersubsetsAand
notAwouldgivethesameplacement(bysettingAtobethesetofballsofeitherurn).Thus
thenumberofplacementsis(2n2)/2=2n11

Q:a)Supposewehaveatrainingsetof10,000spammessagesand5000nonspam
messages.Thewordenhancementappearsin1500spammessagesand20nonspam
messages.Startingwithatheuniformpriorwhatistheposteriorprobabilitythata
messageisspamgiventhatitcontainsthewordenhancement.
A:a)p(enhancement|Spam)=1500/10000=.15,p(enhancement|notSpam)=20/5000=
.004,p(Spam|enhancement)=(.15*.4)/(.15*.5+.004*.5)=.150/.154=.974or97.4%

KnowhowtodoaBayestheoremprobleminvolvingspamfilters(1,2,or3filterwords).

Moreexamplescanbefoundonthehomeworkproblems,whichmightbeworthcopying
ontoyourformulassheet.

16

IDEASFORSTUFFTOCOPYDOWN

Best/worstcasesforalgorithms
Formulas
TableofStirlingnumbersofthesecondkind
Homeworkproblems
Typedupvocab(fromexam2)
WarshallexplainedwellbyLuluTang

17

S-ar putea să vă placă și