33
 Modern Applications using Discrete Mathematical Structures Unit 9 Sikkim Manipal University Page No: 192 Unit 9 Trees and Algorithms Structure 9.1 Introduction Objectives 9.2 Characteri zation of Trees 9.3 Rooted Trees and Applications 9.4 Spanning Trees 9.5 Algorithms for Spanning Trees Self Assessment Questions 9.6 Summary 9.7 Terminal Questio ns 9.7 Answers 9.1 Introduction Trees are extensively used as models in area like computer science, chemistry and i n search proc edures. These are also useful in desi gn of wide range o f algorithms. In this unit we pr ovide some characterizations of trees with suitabl e illustrations. We also introduced a sp ecial type of trees namely roote d trees an d binary trees. We give few applic ations of the se trees. The spanning trees are introduced and the algorithms for spanning trees are also obtained.  Objectives At the end of the unit the student must be able to i) Study the connected graphs without circuits (called trees) and some properties ii) Characteri ze a tree of a connected graph. iii) Know few applications of trees i n realm. iv) Know rooted trees and some of its applications v) Analyze and write algorithms for spanni ng tree s.

MC0063(B)-unit9-fi

Embed Size (px)

DESCRIPTION

Fil

Citation preview

  • ModernApplicationsusingDiscreteMathematicalStructures Unit9

    SikkimManipalUniversity PageNo:192

    Unit9 TreesandAlgorithmsStructure

    9.1 Introduction

    Objectives

    9.2 CharacterizationofTrees

    9.3 RootedTreesandApplications

    9.4 SpanningTrees

    9.5 AlgorithmsforSpanningTrees

    SelfAssessmentQuestions

    9.6 Summary

    9.7 TerminalQuestions

    9.7 Answers

    9.1Introduction

    Trees are extensively used as models in area like computer science,

    chemistry and in search procedures. These are also useful in design of

    widerangeofalgorithms.Inthisunitweprovidesomecharacterizationsof

    treeswithsuitable illustrations. Wealso introducedaspecial typeof trees

    namely rooted treesand binary trees. Wegive fewapplicationsof these

    trees.Thespanning treesare introducedand thealgorithmsforspanning

    treesarealsoobtained.

    Objectives

    Attheendoftheunitthestudentmustbeableto

    i) Study the connected graphs without circuits (called trees) and some

    properties

    ii) Characterizeatreeofaconnectedgraph.

    iii) Knowfewapplicationsoftreesinrealm.

    iv) Knowrootedtreesandsomeofitsapplications

    v) Analyzeandwritealgorithmsforspanningtrees.

  • ModernApplicationsusingDiscreteMathematicalStructures Unit9

    SikkimManipalUniversity PageNo:193

    9.2CharacterizationofTrees

    Theconcept ofa treeplaysa vital role in the theoryof graphs.Firstwe

    introduce the concept of tree, study its properties and some of its

    applications.Later,we introduce theconceptof spanning tree,andstudy

    therelationshipsamongcircuitsandtrees.

    9.2.1Definition

    Aconnectedgraphwithoutcircuits iscalleda tree.Acollectionof trees is

    calledaforest.

    9.2.2Example

    Treeswithone,two,threeandfourverticesaregivenintheFig.9.2.2.

    9.2.3Observations

    i) Atreecontainsatleastonevertex.

    ii) Atreewithoutanyedgeisreferredtoasanulltree.

    iii) Thetreesconsideredarefiniteandatreeisalwaysasimplegraph.

    9.2.4Somenaturalexamples

    i) Thelistoftheancestorsofafamilymayberepresentedbyatree.This

    treereferredtoasafamilytree.

    ii) Ariverwithitstributariesandsubtributariesmayberepresentedbya

    tree.Thistreeisreferredtoasarivertree.

    iii) Thesortingofmailaccordingtozipcodearedoneaccordingtoatree.

    Thistreeiscalleddecisiontree(or)sortingtree.

    Fig.9.2.2

  • ModernApplicationsusingDiscreteMathematicalStructures Unit9

    SikkimManipalUniversity PageNo:194

    9.2.5Theorem

    Inatree T,thereisoneandonlyonepathbetweeneverypairofvertices.

    Proof:Suppose T isatree.

    Bydefinition, T isaconnectedgraphandcontainsnocircuits.

    Since T isconnected,thereexistsatleastonepathbetweeneverypairof

    verticesin T.Supposethatbetweentwovertices a and b of T,there

    are two distinct paths. Now, the union of these two paths will contain a

    circuitin T,acontradiction(since Tcontainsnocircuits).

    Thisshowsthatthereexistsoneandonlyonepathbetweenagivenpairof

    verticesin T.

    9.2.6Theorem:

    Ifthereisoneandonlyonepathbetweeneverypairofverticesin G,thenG

    isatree.

    Proof:Assume that there isoneandonlyonepathbetweeneverypairof

    verticesinagraphG.ClearlyG isconnected.

    Ifpossiblesupposethat G containsacircuit.Thenthereisatleastonepair

    ofvertices a,b suchthattherearetwodistinctpathsbetween a and b,

    whichisacontradictiontoourassumption.So G containsnocircuits.Thus

    Gisatree.

    9.2.7Theorem:

    Atree G withnverticeshas(n1)edges.

    Proof:(Useinductiononn):

    Step(i): If n =1,then G containsonlyonevertexandnoedge.

    Sothenumberofedgesin G is n1=11=0.

    Step(ii):Inductionhypothesis:Thestatementistrueforalltreeswithless

    thannvertices.

  • ModernApplicationsusingDiscreteMathematicalStructures Unit9

    SikkimManipalUniversity PageNo:195

    Step(iii):Letusconsideratreewithnvertices.

    Letekbeanyedgein T whoseendverticesare vi and vj.

    Since T isatree,byTheorem9.2.6, there isnootherpathbetween vi

    and vj.

    Sobyremoving ek from T,wegetadisconnectedgraph.

    Also, Tek consistsofexactlytwocomponents(say T1 and T2).

    Since T isatree,therewerenocircuitsin T andsotherewerenocircuits

    inT1andT2.ThereforeT1 and T2 arealsotrees.

    It isclearthat |V(T1)|+ |V(T2)| = |V(T)| where V(T) denotesthesetof

    verticesin T.

    Also|V(T1)|and|V(T2)|arelessthan n.

    Thereforebytheinductionhypothesis,wehave

    |E(T1)|=|V(T1)|1and|E(T2)|=|V(T2)|1.

    Now|E(T)|1=|E(T1)|+|E(T2)|=|V(T1)|1+|V(T2)|1

    Thus|E(T)|=|V(T1)|+|V(T2)|1=|V(T)|1= n1.

    9.2.8Theorem:

    Anyconnectedgraphwithnverticesand n1edgesisatree.

    Proof:LetGbeaconnectedgraphwith n verticesand n1edges.

    Itsufficestoshowthat Gcontainsnocircuits.

    If possible, suppose that G contains a circuit. Let e be an edge in that

    circuit.

    Since einacircuit,wehavethat Ge isstillconnected.

    Now Ge isconnectedwithnvertices,andso|Ge| n1edges,a

    contradiction(tothefactthat Ge containonly(n2)edges,bytheorem

    9.2.7).SoGcontainsnocircuits.Therefore G isatree.

  • ModernApplicationsusingDiscreteMathematicalStructures Unit9

    SikkimManipalUniversity PageNo:196

    9.2.9Definition:

    Aconnectedgraphissaidtobeminimallyconnected if theremovalofany

    oneedgefromthegraphprovidesadisconnectedgraph.

    9.2.10Example

    i) GraphgiveninFig.9.2.10Aisnotminimallyconnected.

    ii) GraphgiveninFig.9.2.10B isminimallyconnected.

    iii) Everycircuitisnotminimallyconnected.

    iv) Everytreeisminimallyconnected.

    9.2.11Problem

    Agraph G isatreeifandonlyifitisminimallyconnected.

    Solution: Part (i): Assume that G is a tree. If G is not minimally

    connected,thenthereexistsanedgeesuchthat Ge isconnected.

    Thatis,e isinsomecircuit,whichimplies G isnotatree,acontradiction.

    HenceG minimallyconnected.

    Part(ii):Supposethat G isminimallyconnected.

    IfG containsacircuit,thenbyremovingoneoftheedgesinthecircuit,we

    getaconnectedgraph,acontradiction(tothesupposition).Thisshowsthat

    G containsnocircuits.ThusG isatree.

    Observation:To join ngivendistinctpoints, theminimumnumberof line

    segmentsneededis n1.

    Fig.9.2.10A Fig.9.2.10B

  • ModernApplicationsusingDiscreteMathematicalStructures Unit9

    SikkimManipalUniversity PageNo:197

    9.2.12Problem::

    Ifagraph G contains n vertices, n1edgesandnocircuits,then G isa

    connectedgraph.

    Proof: Let G beagraphwithnvertices, n 1 edgesandcontainsnocircuits.

    Supposethat(inacontraryway)Gisdisconnected.

    G consistsoftwoormorecircuitlesscomponents(say, g1, g2,, gk)(herek 2).

    Chooseavertexvi ingi,for1 i k.

    Addnewedges e1,e2,,ek1 where ei= 1 +iivv togetanewgraph G*.

    Itisclearthat G* containsnocircuitsandconnected,andso G* isatree.

    Now G*containsnverticesand

    (n 1)+ (k 1)= (n+k 2) n edges,acontradiction (sincea tree

    contains(n1)edges). Thisshowsthat G isconnected.

    9.2.13CharacterizationforTree

    Foragivengraph G,thefollowingconditionsareequivalent:

    i) G isconnectedandiscircuitless,

    ii) G isconnectedandhas n1edges,

    iii) G iscircuitlessandhas n1edges,

    iv) Thereisexactlyonepathbetweeneverypairofverticesin G,

    v) G isaminimallyconnectedgraph,

    vi) G isatree.

    Proof: (i) (vi)isclear.

    (vi) (ii)and(iii):Theorem9.2.7

    (ii) (vi):Theorem9.2.8

    (iii) (vi):Theorem9.2.12

    (iv) (vi):Theorems9.2.5and9.2.6

    (v) (vi):Problem9.2.11.

  • ModernApplicationsusingDiscreteMathematicalStructures Unit9

    SikkimManipalUniversity PageNo:198

    9.2.14Problem

    If G is a tree (withminimum two vertices), then there exists at least two

    pendantvertices.

    Proof:LetGbeatreewith|V| 2.

    Let v0e1v1e2v2e3 . . . vn1envn bea longestpath in G (since G isfinite

    graph,itispossibletofindalongestpath).

    Nowweshowthat d(v0)=1=d(vn).

    If d(v0)>1,thenthereexistsatleastoneedge e withendpoint v0 such

    that e e1.If e {e1,e2,en},then e = ei forsome i 1.

    Soeither vi1= v0 or vi= v0 v0 repeatedinthepath,acontradiction.

    Therefore e {e1, e2, e3 ,en}andsoe,e1, e2, e3 ,en isapathof

    length n+1,acontradiction.

    Thus d(v0)=1.Inasimilarway,wecanshowthat d(vn)=1.Hence

    v0,vn aretwopendantvertices.

    9.2.15Definition

    LetGbeaconnectedgraph.Thedistancebetweentwoverticesvandu is

    denotedby d(v,u)andisdefinedasthelengthoftheshortestpath[that

    is, the number of edges in the shortest path] between v and u. In a

    connectedgraph,wecanfindthedistancebetweenanytwogivenvertices.

    9.2.16Example

    i) ConsidertheconnectedgraphgiveninFig.9.2.16A.Heresomeofthe

    pathsbetween v1 and v2 are(a,e),(a,b,f), (b,c,e),(b,f),(b,

    g,h),(b,g,i,j),(b,g,i,k).Heretherearetwoshortestpaths(a,e)

    and(b,f)eachoflength2.Hence d(v1,v2 )=2.

  • ModernApplicationsusingDiscreteMathematicalStructures Unit9

    SikkimManipalUniversity PageNo:199

    ii) ConsiderthetreegivenintheFig.9.2.16B.Here d(a,b)=1, d(a,c)=

    2, d(a,d)=2, d(b,d)=1.

    9.3RootedTreesandApplications

    Rootedtreesareextensivelyusedinthecomputersearchmethods,binary

    identificationproblems,andvariablelengthbinarycodes.

    9.3.1Definition

    Atreeinwhichonevertex(calledtheroot)isdistinguishedfromalltheother

    vertices,iscalledarootedtree.Inarootedtree,therootisgenerallymarked

    inasmalltriangle(orsmallcircle).

    9.3.2Example

    Distinctrootedtreeswithfourvertices,weregiveninFig.9.3.2

    Generally, the term treemeans treeswithoutany root.However theyare

    sometimescalledfreetrees(or) nonrootedtrees.Averityofrootedtrees

    (calledtheBinaryrootedtrees)isofparticularinterest.

    j

    b

    d

    c

    v2

    v1i

    h

    k

    g

    e

    a

    fFig.9.2.16A

    a

    b cdFig9.2.16B

    Fig.9.3.2

  • ModernApplicationsusingDiscreteMathematicalStructures Unit9

    SikkimManipalUniversity PageNo:200

    9.3.3Definition:

    A tree in which there is exactly one vertex of degree 2, and all other

    remainingverticesareofdegreeoneorthree,iscalledabinarytree.

    i) TheFig.9.3.3representsabinarytree(sincetheonlyvertexv1isof

    degree2,andallotherverticesareofdegreeeither1or3).

    ii) Thevertexofdegree2 (that is, v1) is distinct fromallothervertices,

    thisvertex v1 istheroot.

    iii) Inabinary tree, the vertexwithdegree2 servesasa root.Soevery

    binarytreeisarootedtree.

    9.3.4 PropertiesofBinarytrees

    Property(i):Thenumberofvertices n,inabinarytreeisalwaysodd.

    Property(ii):Thenumberofpendentverticesis21n+ .

    Property(iii):Numberofverticesofdegree3is =np1=n(2

    1n+ )

    1=2

    3n- .

    v3

    v2

    v4

    v6v8

    v5

    v11

    v7

    v1

    v10

    v13

    v9

    v12

    Fig9.3.3

  • ModernApplicationsusingDiscreteMathematicalStructures Unit9

    SikkimManipalUniversity PageNo:201

    9.3.5Example

    InthegraphgiveninFig.9.3.3,wehavethat n=13, p=2

    1n+ =

    2

    113+

    =214

    =7.Thereforenumberofverticesofdegree3is2

    3n- =

    2

    313- =5.

    9.3.6Definition:

    Anonpendentvertexinatreeiscalledaninternalvertex.

    Observation:

    i) ThenumberofinternalverticesinaBinarytreeis

    21n- =(p1)where p =thenumberofpendentvertices.

    ii) InthebinarytreegiveninFig.9.3.3,theinternalverticesarev1,v3,v4,

    v5,v6,v9.Theseare6(=71= p1)innumber.

    9.3.7Definition:

    Letv beavertexinabinarytree.Thenv issaidtobeatlevel I ifv isata

    distanceofI fromtheroot.

    9.3.8 Example:

    i) A13vertex,4levelbinarytreewasgiveninFig.9.3.8.

    level1

    level4

    level0

    level2

    level3

    Fig.9.3.8

  • ModernApplicationsusingDiscreteMathematicalStructures Unit9

    SikkimManipalUniversity PageNo:202

    Here the number of vertices at levels 0, 1, 2, 3, 4 are 1, 2, 2, 4 and

    4respectively.

    9.3.9Definition:

    Thesumofpath lengthsfromthe root toallpendentvertices iscalled the

    pathlength(or) externalpathlengthofatree.

    9.3.10Example:

    i) ThepathlengthofthebinarytreegiveninFig.9.3.8is:

    1+3+3+4+4+4+4=23.

    ii) IntheFigures9.3.10A and B,therearetwo11vertexbinarytrees.

    Thepathlengthofgraph(fig.9.3.10A):2+2+3+3+3+3=16.

    Thepathlengthofgraph(fig.9.3.10B):1+2+3+4+5+5=20.

    level5

    level0

    level1

    level2

    level4

    level3

    Fig9.3.10B

    level 1

    level 0

    level 2

    level 3

    Fig9.3.10A

  • ModernApplicationsusingDiscreteMathematicalStructures Unit9

    SikkimManipalUniversity PageNo:203

    9.3.11Searchprocedures

    Eachvertexofabinarytreerepresentsatestwithtwopossibleoutcomes.

    Westartattheroot.Theoutcomeofthetestattherootsendsustooneof

    thetwoverticesat thenext level,wherefurthertestsaremadeandsoon.

    Reachingaspecifiedpendentvertex(thatvertexwhichrepresentsthegoal

    ofthesearch),terminatesthesearch.

    Forsuchsearchprocedures,itisoftenimportanttoconstructabinarytreein

    which,foragivennumberofvertices n,thevertexfrothiestfromtherootin

    asclosetotherootaspossible.

    i) Therecanbeonlyonevertex(theroot)atlevel0.Numberofvertices

    atleveloneisatmost2.Numberofverticesatleveltwoisatmost22

    andsoon.So themaximumnumberofverticespossible inak level

    binarytreeis20 +21 +22 +...+2k.

    So n 20 +21 +22 +...+2k

    ii) Themaximumnumberamongthelevelsoftheverticesinabinarytree

    iscalledheightofthetree.Soheight=max{levelofavertex v /

    v V}.

    Thisheightisdenotedby lmax.

    iii) Toconstructabinarytreeforagiven n suchthatthefarthestvertex

    isasfaraspossiblefromtheroot,wemusthaveexactlytwoverticesat

    eachlevel,exceptatthe0level.Somaxlmax =2

    1n- .

    9.3.12AnApplication:(CokeMachineProblem)

    Supposethat there isa.Themachine is tohaveasequenceof tests (for

    example, it should be capable of identifying the coin that is put into the

    machine).We suppose that five rupees coin, two rupees coin, one rupee

    coinandfiftypaisecoincangothroughtheslot.Sothemachinecanidentify

    only these four coins.Every coin put in, is tobe testedby themachine.

    Each test got the effect of partitioning the coins into two complementary

  • ModernApplicationsusingDiscreteMathematicalStructures Unit9

    SikkimManipalUniversity PageNo:204

    sets.[Supposeacoinisputintothemachine.Itshouldtestwhetherthecoin

    isfiverupeecoin.Ifitisnotafiverupeescoin,thenitshouldtestwhetheritis

    atworupeescoinandsoon].Wesupposethetimetakenforeachtestis.

    TestPattern1:Onetypeof testingpatternwasshowninGraph(i), given

    inFig.9.3.12A.

    Supposethestatisticaldatatellsthat

    w1 =probabilityofputtingaRs 5coin=0.5

    w2 =probabilityofputtingaRs 2coin=0.2

    w3 =probabilityofputtingaRs 1coin=0.2

    w4 =probabilityofputtingaRs 0.5coin=0.1

    Now

    )v(lw ii = )v(lw 11 + )v(lw 22 + )v(lw 33 + )v(lw 44 =(0.5)(1) +(0.2)(2)+(0.2)(3)+(0.1)(4)=1.9

    Soexpectedtimetobetakenbythemachinefortestingonecoinis1.9t.

    Thus if themachinefollows (for its testingpattern) thebinary treegiven in

    Graph(a),thentheexpectedtimefortestingonecoinisequalto1.9t.

    Test Pattern2: Another type of testing pattern was given in the

    Fig.9.3.12B.

    Rs0.5coin

    Rs1coin

    Rs5coin

    v30.2

    v1

    v2

    v40.1

    0.2

    0.5

    notRs1coin

    notRs0.5coin

    notRs2coin

    notRs5coin

    Fig9.3.12A

    Rs2coin

  • ModernApplicationsusingDiscreteMathematicalStructures Unit9

    SikkimManipalUniversity PageNo:205

    )v(lw ii = )v(lw 11 + )v(lw 22 + )v(lw 33 + )v(lw 44 =(0.5)(2)+(0.2)(2)+(0.2)(2)+(0.1)(2)=2.0

    Sohere,theexpectedtimetobetakenbythemachinefortestingonecoin

    is2t.

    Thus if themachinefollows (for its testingpattern) thebinary treegiven in

    graph9.3.12Bthentheexpectedtimefortestingonecoinis2t.

    9.3.13Definition:

    Agraph G inwhicheveryvertexisassignedauniquenameorlabel(that

    is,notwoverticeshavethesamelabel)iscalledalabeledgraph.Otherwise

    issaidtobeanunlabeledgraph.

    9.3.14Example:

    Ifn =4,thenthereare16trees,shownbelow.

    TakethevertexsetV ={A,B,C,D}.

    v1 v2 v3 v40.10.2

    0.20.5

    Rs5or Rs2

    Rs5 Rs2Rs1

    Rs0.50

    Rs1or Rs0.50

    Fig9.3.12B

  • ModernApplicationsusingDiscreteMathematicalStructures Unit9

    SikkimManipalUniversity PageNo:206

    Thefollowingarethe16treesoffourlabeledvertices

    B

    C

    A

    D

    A

    DC

    BA

    C D

    B

    C

    BA

    D

    A B

    DC DC

    BA

    A B

    C D

    A B

    C D

    A

    C

    B

    D

    B

    D

    A

    C

    A B

    C D

    A B

    C DA B

    C D

    A B

    CD

    A B

    C D

    A B

    CD

  • ModernApplicationsusingDiscreteMathematicalStructures Unit9

    SikkimManipalUniversity PageNo:207

    Observations:

    i) Whenwecountthenumberofdifferentgraphs,thedistinctionbetween

    labeledandunlabeledgraphsisimportant.

    ii) ConsiderthegraphsinExample9.3.14,the5th,6th,7th and8th,are

    counted as four different trees (even through they are isomorphic),

    becausetheyarelabeled.Ifthereisnodistinctionbetween A,B,C,

    D,thenthesefourtreescountedasone.

    iii) A careful inspectionof the 16 graphswill reveal that thenumberof

    unlabeledtreeswithfourvertices(nodistinctionmadebetween A, B,

    C, D),istwo.Thesetwographswere.

    9.3.15CayleysTheorem:

    Thenumberoflabeledtreeswith n vertices(n 2)is n(n2).

    9.3.16 Example:Part(i):Suppose thegiven tree is T1 (Fig. 9.3.16A). a1 = thependentvertexwithsmallestlabel.So a1 isthevertex2.Now b1 =1.Afterremoving a1 andtheedge(a1,b1), intheremaininggraph,a2 = thependentvertexwithsmallestindex=4.

    a1

    1

    2

    6

    5

    3

    79

    4

    8

    Fig.9.3.16A

  • ModernApplicationsusingDiscreteMathematicalStructures Unit9

    SikkimManipalUniversity PageNo:208

    Now b2 =1. a3=1, b3=3, a4 =3, b4 =5, a5 =6, b5 =5, a6 =

    7, b6=5, a7=5, b7=9.

    Thereforewehavethesequence(b1, b2,b3, b4,b5, b6,b7)=(1,1,3,5,5,5,9).

    Part(ii): (converseofpart(i)):wehavetoconstructatreewith n =9vertices.

    Consider1,2,3,4,5,6,7,8,9.........(i)

    Given(n2)tupleis(1,1,3,5,5,5,9).......(ii)

    Observethesequencein (ii). First join1and3, 3and5, 5and9.

    Thenweget thegraphgiven inFig.9.3.16B. Nowthe leastnumber in (i)

    whichisnotin(ii),is2.Sowejoin2and1.Nextleastin(i)whichisnotin(ii)

    is4.Sowejoin4and1.Cancel3in(i)and3in(ii).Alsocancel5in(i)

    and5 in (ii) (in (ii), cancelingofonlyone 5 isallowed).Thenext least

    whichisnotin(ii)is6.Sowejoin5and6.Thenextleastwhichisnotin(ii)

    is 7. So we join 7 and 5. The next least which is not in (ii) is 8. The

    remainingnumberavailablein(ii)is9.Sowejoin8and9.

    13

    5

    9

    Fig9.3.16B

    8

    1

    2

    6

    5

    3

    7

    9

    4

    Fig9.3.16C

  • ModernApplicationsusingDiscreteMathematicalStructures Unit9

    SikkimManipalUniversity PageNo:209

    Nowwewillstudythetreeasasubgraphofanothergraph.Agivengraph

    mayhavenumeroussubgraphs.If e isthenumberofedgesin G,then

    there are 2e distinct subgraphs are possible. Obviously some of these

    subgraphs will be trees. Out of these trees we particularly interested in

    certaintypeoftrees,calledspanningtrees.

    9.4SpanningTrees

    9.4.1Definition:

    Atree T issaidtobeaspanningtreeofaconnectedgraph G if T isa

    subgraphof G and T containsalltheverticesof G.

    9.4.2 Example: Consider the graph G given in Fig. 9.4.2 A. Graph T

    (giveninFig.9.4.2B)isaspanningtreeof G.

    v3

    v1

    v4v2

    v5

    v6

    v7b2

    b3

    b1

    b6

    b4

    c5

    c6

    c8

    c7

    c1

    b5

    c2

    c3

    c4

    Fig.9.4.2AGraphG

    v7

    v1

    v2

    v3v4

    v5

    v6

    Fig.9.4.3BGraphT

  • ModernApplicationsusingDiscreteMathematicalStructures Unit9

    SikkimManipalUniversity PageNo:210

    Observations:

    i) Spanning treesare the largest (with themaximumnumberofedges)

    treesamongalltreesinG.Spanningtreeisalsocalledamaximaltree

    subgraphormaximaltreeof G.

    ii) Spanningisdefinedonlyforaconnectedgraph(sinceatreeisalways

    connected).

    iii) Eachcomponentofadisconnectedgraph,doeshaveaspanningtree.

    Thus a disconnected graph with k components contains a spanning

    forestconsistingof k spanningtrees.

    9.4.3Theorem:

    Everyconnectedgraphhasatleastonespanningtree.

    9.4.4Definition:

    i) Anedgeinaspanningtree T iscalledabranchof T.

    ii) Anedgeof G that isnot inagivenspanningtree T iscalleda

    chord.Inelectricalengineering,chordsometimesreferredtoastieora

    link.

    iii) Let T beanyspanningtreeofaconnectedgraph G,and T1 isthe

    complementof T in G.Theneachedgein T iscalledabranch(with

    respectto T),andthesetofalledgesin T iscalledthebranchset.

    Eachedgein T1 iscalledachord(withrespectto T),andthesetof

    edges in T1 is called thechord set (or)Tieset. T1 is calledas the

    cotree.WemaywriteT insteadof T1.

    Observation:

    i) Branchesandchordsaredefinedonlywithrespecttoagivenspanning

    tree.

    ii) Anedgethatisabranchwithrespecttoonespanningtree T1 (of G)

    maybeachordwithrespecttoanotherspanningtree T2.

  • ModernApplicationsusingDiscreteMathematicalStructures Unit9

    SikkimManipalUniversity PageNo:211

    9.4.5Theorem:

    Withrespecttoanyofitsspanningtrees,aconnectedgraphofnvertices

    andeedgeshasn1treebranchesand en+1chords.

    Proof:LetG beanyconnectedgraphonnverticesand e edges.Let

    T beanyspanningtreein G. Sinceeveryspanningtreeof G contains

    allverticesof G,wehavethat|V(T)|= n andso|E(T)|=|V(T)|1= n

    1.Sinceeveryedgeofaspanningtree T iscalledabranchofT,wehave

    that Gcontains n1branches.SincethenumberofedgesinG is e,we

    havethatthenumberofchordsof Tis e(n1)=en+1.

    9.4.6Example:

    Thereisaformconsistingofsixwalledplotsoflandasshowninthefigand

    theseplotsarefullofwater.Wefindtheminimumnumberofwallsaretobe

    brokensothatallthewatercanbedrainedout.

    Consider the wall joints as vertices, and walls as edges. Then we can

    consider itasagraph. Inthisgraphthenumberofvertices is n=10,and

    thenumberofedgesis e=15.

    Fig.9.4.6

    If thereexistsacircuit, thenthewater inside thecircuitcannotbedrained

    out.Sowehavetoremoveminimumnumberofedgessothatthegraphdo

    not contain circuits. Tohave this, weshouldhave a spanning treewith

    (n1)edges.Hencewehavetobreak e(n1)= en+1=1510

    +1=6edges(walls)sothatallthewatercanbedrainedout.

    If we add an edge between any two vertices of a tree, then a circuit is

    created. This is because, therealreadyexistsonepathbetweenany two

  • ModernApplicationsusingDiscreteMathematicalStructures Unit9

    SikkimManipalUniversity PageNo:212

    verticesofa tree, addinganedge inbetween, createsanadditionalpath,

    andhenceacircuit.Theconceptofafundamentalcircuithasanenormous

    significanceinelectricalnetworkanalysis.

    9.4.7Definition:

    Let T beany spanning treeof a connectedgraph G.Addinganyone

    chordto T willcreateexactlyonecircuit.Suchacircuitformedbyaddinga

    chordtoaspanningtree,iscalledafundamentalcircuit.

    9.4.8 Example: Consider the graph G (given in Fig. 9.4.8A), and its

    spanningtree T(giveninFig.9.4.8B)of G.Now,ifweaddthechord c1

    to T,wegetacircuitb1b2b3b5c1whichiscalledasfundamentalcircuit

    (giveninFig.9.4.8C).

    9.4.9Problem:AconnectedgraphG isatree ifandonly if addingan

    edgebetweenanytwoverticesin G createsexactlyonecircuit.

    Prof:SupposeGisatrue.Supposeweaddanewedge e= uv between

    twovertices u and v in G.Since G isatree, G containsapathfrom u

    and v.Byjoiningthisnewedge,therecreatesexactlyonecircuit.

    c7c5

    c6

    c8

    c4c3

    c1

    b2

    b1

    b4

    b5b3b6

    c2

    Fig9.4.8AGraphG

    b1

    b2

    b3

    b4

    b6b5

    Fig.9.4.8BSpanningtreeT

    c1b1

    b2

    b3

    b4

    b6b5

    Fig.9.4.8CFundamentalcircuitF

  • ModernApplicationsusingDiscreteMathematicalStructures Unit9

    SikkimManipalUniversity PageNo:213

    Converse: Supposetheconversehypothesis.SupposeGcontainsacircuit:

    v1e1v2e2env1.Byaddingnewedge e1= v1v2 ,wegettwocircuits v1e1

    v2e1v1 and v1e1v2e2v3env1, acontradiction.So G containsnocircuits.

    NowweshowthatGisconnected.Let u,v betwoverticesinG.Adda

    newedge e*= uv toG.Thenbyconversehypothesis,therecreatesa

    circuit.Supposethecircuitis ue*e1v1e2v2.eku.Now ve1v1e2v2ekuis

    apathinbetweenv and u.Thisshowsthatthegraph G isconnected.

    Hence G isatree.

    9.4.10Problem:

    ForatreeG isatree,byaddinganewedgebetweenanytwoverticesin G

    creates,exactlyonefundamentalcircuit.

    9.5AlgorithmsforSpanningTrees

    9.5.1Definition:

    i) AgraphG issaidtobeaweightedgraphifalltheedgese ofG were

    assignedbyacorrespondingrealnumber w(e)(readastheweightofe).

    ii) Let T beanyspanningtreeofaconnectedgraph G.thentheweight

    of T,thatis, w(T)isdefinedasthesumofweightsofallbranches

    inT.

    iii) Aspanningtreewiththesmallestweightinaweightedgraphiscalleda

    shortestspanningtree(or)minimalspanningtree(or)shortestdistance

    spanningtree.

    Observation:

    i) Different spanning treesofG mayhavedifferentweights.Amongall

    the spanning trees of G, one with the smallestweight is of practical

    significance.

    ii) Let G beagraphon n verticesinwhicheveryedgehasaunitweight.

    Thenallthespanningtreeshavethesameweightof(n1)units.

  • ModernApplicationsusingDiscreteMathematicalStructures Unit9

    SikkimManipalUniversity PageNo:214

    9.5.2AnApplication:

    i) Suppose that we are to connect n cities v1, v2,, vn through a

    networkofroads.

    ii) ThecostCij ofbuildingadirectroadbetween vi and vj isgivenfor

    pairsofcities vi and vj whereroadscanbebuilt.(Notethatthere

    maybepairsofcitiesbetweenwhichnodirectroadcanbebuilt).

    iii) Nowtheproblemis:Tofindtheleastexpensivenetworkthatconnects

    all n cities together. That is, to find a shortest spanning tree in a

    connectedweightedgraphofnvertices.

    iv) Heretheconnectednetworkmustbeatree.Otherwise,wecanalways

    removesomeedgesandget a connectedgraphwith smallerweight.

    Thus the problem of connecting n cities with a least expensive

    network is the problem of finding a shortest spanning tree in a

    connectedweightedgraphofnvertices.

    9.5.3Problem: Let T bespanning tree (ofaweightedconnectedgraph

    G).Thenthefollowingconditionsareequivalent:

    i) T isashortestspanningtree(of G)and

    ii) Thereexistsnospanningtree(of G)atadistanceonefrom T whose

    weightissmallerthanthatof T.

    9.5.4KruskalAlgorithm:(findingshortestspanningtree)

    Step(i):ListalltheedgesofGinorderofnondecreasingweight.Nowwe

    selectanedgee1ofGsuchthat w(e1)isassmallaspossibleande1isnot

    aloop.

    Step(ii):Selectnextsmallestedgefrom thesetofall remainingedgesof

    G such that the selectededgedonot formacircuitwith theedges that

    havealreadybeenchosen.

    Step(iii):Wecontinuethisprocessof takingsmallestedgesamongthose

    notalreadychosen,providednocircuitisformedwiththose,thathavebeen

    chosenalready.

  • ModernApplicationsusingDiscreteMathematicalStructures Unit9

    SikkimManipalUniversity PageNo:215

    [If edges e1, e2, , ei have been chosen, then chose ei+1 from E

    {e1,e2,,ei}insuchwaythatgraphwith{e1,e2,,ei+1}isacyclicand

    w(ei+1)isassmallaspossible].

    Step(iv):Ifagraph G hasnvertices,thenwewillstopthisprocess

    afterchoosing n1edges.Theseedgesformasubgraph T,whichisnot

    cyclic.(Thus T isashortestspanningtreeof G).

    9.5.5 Example:

    ConsidertheconnectedweightedgraphG.Listing:

    )1(AB , )1(AD , )2(BC , )2(CD , )3(AE , )3(FE , )3(AF , )4(EB ,

    )4(AC , )4(ED , )5(FD .

    Step(i):Consideraspanningsubgraphof G withoutedges.

    Select theedgewithminimumweight. For thisgraph theedgeAB isof

    minimumweight.

    Add this edge e1 = AB to the spanning null graph. Then we get the

    subgraphG1.

    Step(ii):Selecttheedge e2 suchthat G1+e2 containsnocircuitsand e2

    gotminimumweightamongsuchedges.Forthisgraph e2=AD.Nowadd

    e2 = ADto G1.ThenwegetthesubgraphG2.

    Step(iii):Wecontinuethisprocess.Forthisexample,weadd e3 = BCto

    G2 togetthesubgraphG3.

  • ModernApplicationsusingDiscreteMathematicalStructures Unit9

    SikkimManipalUniversity PageNo:216

    Nextweadd e4 =AE to G3 toget the subgraphG4. Nextweadd

    e5 = EF toget the subgraphG5. Ifweaddanyedge to G5, weget a

    circuit.Sotheprocessistobestoppedhere.

    A

    B

    E

    1F

    12

    Graph G5 or T

    D C

    33

    F

    A

    B

    E

    1

    12

    SubgraphG4

    DC

    3

    F

    E

    A

    B

    1

    12

    SubgraphG3

    D CC

    A

    B

    1

    1 SubgraphG2

    D

    F

    E

    C

    2

    5 1

    D

    4

    A

    B

    E

    F

    1

    23 4

    3

    3

    4

    GraphG

    A

    B1

    SubgraphG1

    F

    E

    D

    C

  • ModernApplicationsusingDiscreteMathematicalStructures Unit9

    SikkimManipalUniversity PageNo:217

    Thefinalgraph G5 isasubgraphof G andcontainsnocycles.Thus G5

    isashortestspanningtreeofG.Write T = G5.Now w(T)=sumofthe

    weightsofthebranchesin T =1+2+1+3+3= 10

    9.5.6Prim'sAlgorithm:

    Step(i):Chooseanyvertex v1 in G.

    Step(ii):Chooseanedgee1=v1v2 of G suchthat v1 v2ande1has

    the smallest weight among the edges of G incident with v1 . [For

    convenience,wecanformatableindicatingtheweights].

    Step(iii): Iftheedges e1,e2,,ei havebeenalreadychoseninvolving

    endvertices v1,v2,,vi+1 ,chooseanedgeei+1,where ei+1=vjvk with

    vj {v1,v2,,vi+1},and vk {v1,,vi+1}suchthat ei+1 hasthesmallest

    weightamongtheedgesof G withpreciselyoneendin{v1,v2,,vi+1}.

    Notethatafteraddingei+1 thegraphshouldbeacyclic.

    Step(iv): The process will stop after choosing the n1 edges (otherwise,

    repeatthestep(iii)).

    9.5.7Illustration:Considertheconnectedweightedgraph G.Letusstart

    with C.

    5 1

    D

    4 2

    A

    C

    B

    E

    F

    1

    23 4

    33

    4

    GivengraphG GraphG1D

    A

    F

    E

    B

    C

  • ModernApplicationsusingDiscreteMathematicalStructures Unit9

    SikkimManipalUniversity PageNo:218

    Step(i):Takethespanningnullsubgraph G0 of G.

    Letuschoosev1=A.TheedgeAD isincidenton v1 andhasthesmallest

    weightamongtheedgesincidentonv1=A.WriteG1 =(G0+AD)

    Step(ii): InthisstepweselectDCandwriteG2=(G1+DC).

    Step(iii): Weselect AB andwrite G3=(G2+ AB).

    Step(iv):Weselect AF andwrite G4=(G3+AF).

    Step(v):WeselectFEandwriteG5=(G4+FE).

    Step(vi):Since G5 isaspanningtree,theprocessstopshere.

    Write T = G5.

    Hence,theshortestspanningtreeof G isG5=T.

    Now w(T)=1+1+2+3+3=10.

    A

    B

    E

    1

    F

    1

    2D

    C

    3

    3

    GraphG5=T

    A

    B1

    F

    1

    2D

    C

    3

    GraphG4

    E

    A

    B1

    1

    2D

    C

    GraphG3

    F

    E

    A

    1

    1

    D

    C

    GraphG2

    F

    E

    B

  • ModernApplicationsusingDiscreteMathematicalStructures Unit9

    SikkimManipalUniversity PageNo:219

    SelfAssessmentQuestions:

    1. Drawtreeswith5vertices.

    2. Findallspanningtreesofthegraph.

    3. Findallspanningtreesofthefollowinggrpah.

    4. The complete graph on n vertices Kn: has2 -nn different spanning

    trees.GiveallthespanningtreesofK4.

    5. Findaspanningtreeforeachofthefollowinggraphs

    6. Whichconnectedsimplegraphshaveexactlyonespanningtree?

    o oo

    o

    o

    oo o

    o o

    (i)o

    o

    o

    o o

    o o

    o o

    (ii)

    o

    o

    o

    o

    (iii)

    o

    o

    o o

    o

    (iv)

    o

    o

    ooo

    o

    o o

    o

  • ModernApplicationsusingDiscreteMathematicalStructures Unit9

    SikkimManipalUniversity PageNo:220

    7. UsePrimsAlgorithmtofindaminimalspanningtreeforthefollowing

    weightedgraph.

    9.6Summary:

    Thisunitdealtwithaspecialtypeofgraphsnamelytrees,whichareextensively

    usedsearchproceduresanddesignofcomputeralgorithms.Thereaderwillbe

    abletogetandideatofindthedifferentspanningtreesfromagivenconnected

    graph. Different characterizations and properties of trees were given. Also

    someapplicationsofbinarytreeswerediscussed.Spanningtreeplayavital

    roleinmulticastingoverinternetprotocolnetworks.

    9.7TerminalQuestions

    1. Draw two different binary trees with five vertices having maximum

    numberofleaves.

    2. Drawagraphwiththegivenspecification.Incase,ifnotpossible,then

    explainwhynosuchgraphsexists.

    i) fullbinarytree,fiveinternalvertices

    ii) fullbinarytree,fiveinternalvertices,seventerminalvertices

    iii) fullbinarytree,twelvevertices

    iv) fullbinarytree,ninevertices.

    3. If T1 andT2 aretwospanningtreesinagraph,thenshowthat

    i) Thenumberofedgesin T1 not in T2 isequaltothenumberof

    edgesinT2 notin T1.(ii) d(T1,T2)= d(T2,T1).

    4. LetN(g)=thenumberofedgesinagraph g.Then N(Ti Tj)=the

    8

    A

    3

    6

    47

    2

    46

    8 6

    C

    BE

    D

    o

    o

    o

    o

    o

  • ModernApplicationsusingDiscreteMathematicalStructures Unit9

    SikkimManipalUniversity PageNo:221

    numberofedgesinTi Tj ,and d(Ti,Tj)=Thedistancebetween Ti

    and Tj.Showthat d(T1,T2)=21

    N(T1 T2).

    5. Considerthetableofairlinedistancesinmilesbetweensixoflargestcities

    intheworld:London,Mexico,NewYork,Paris,PekingandTokyo.

    L MC NY Pa Pe T

    L 5558 3469 214 5074 5959

    MC 2090 5725 7753 7035

    NY 3636 6844 6757

    Pa 5120 6053

    Pe 1307

    T

    9.8Answers

    SelfAssessmentQuestions

    1.

    2. The given graph has four vertices and so each spanning tree must

    have41=3edges.Theyare

    o

    o

    o

    o

    o

    oo

    o

    o

    o

    o

    o

    o

    o

    o

    o o

    o

    o

    o

    o o o o o o

    o

    o o

    o

    o

    o

    o

    oo o oo

    o o

    ooo oo o

    o

    L

  • ModernApplicationsusingDiscreteMathematicalStructures Unit9

    SikkimManipalUniversity PageNo:222

    4. K4has244 - =16differentspanningtrees.Eachhavingthreeedges.

    5.

    6. Tree.

    7. ChooseedgesAE,AC,DC,AB.

    3.o o

    o

    o

    oo o o o

    o o o oo

    oo o

    o

    (i) o

    oo

    o

    o

    o

    o

    o

    o

    (ii) o

    oo o

    (iii) o

    o

    o

    o

    o

    (iv)

    o

    o o

    o

    o

    oo

    o

    o

  • ModernApplicationsusingDiscreteMathematicalStructures Unit9

    SikkimManipalUniversity PageNo:223

    TerminalQuestions

    1.

    2.

    ii) Not possible, since any full binary tree with five internal vertices

    havesixterminalvertices,butnotseven.

    iii) No,afullbinarytreehas2k+1verticeswherekisthenumberof

    internalverticesbut2k+1isodd.

    iv)

    3. i) Let T1,T2 betwospanningtrees.Now|E(T1)|= n1=|E(T2)|.

    Suppose T1 contains m edgeswhicharenotin T2.Nowthe

    numberofedgesin T2 notin T1 is

    (n1)[(n1)m]= m.followsfrom(i).

    (i)

    b

    c

    a

    e

    f g

    j k

    o

    o

    oo

    o

    o

    o

    d

    h io o

    o o

    o

    oo

    o

    oo

    o

    oo

    o

    o o

    o o o

    o

    oo

    o

  • ModernApplicationsusingDiscreteMathematicalStructures Unit9

    SikkimManipalUniversity PageNo:224

    4. N(T1 T2)=numberofedgesinT1 T2 =(numberofedgesinT1but

    notinT2)+(numberofedgesinT1butnotinT2)= d(T1,T2)+d(T2,

    T1)=2d(T1,T2) d(T1,T2)=21

    N(T1 T2)

    5.

    214

    2090

    3469

    5074

    1307

    MC

    NY

    T

    L

    Pa

    Pe

    FigB Shortest spanning tree

    5725

    6053

    7753

    6757

    50743469

    5120 3636

    5558

    2090

    5959

    1307

    T

    Pe

    MC

    NY

    Pa

    7035

    6844

    214

    Fig.A