Relations, Partition and Poset

Basic definitions
binary relation, R: X (X), R(x): the relatives of x R: a collection of ordered pairs R XX Symmetric property antisymmetric property reflective property irreflexity property x'R(x) xR(x') x'R(x) xR(x') x R(x) xR(x)

transitive property zR(y), yR(x) zR(x) transitive extension, R1

R1 R, and if (a,b) and (b,c) are in R, then (a,c) is in R1. transitive closure, R* R* = R1 R !

e"uivalent relation and partition e"uivalent relation: reflexive, symmetric and

transitive partition of a set: a division of the elements of a set into disjoint subsets

partition induced by an equivalent relation

refinement: Let 1 (resp. R1) and (resp. R ) be t!o partitions of a set X. "e say that 1 is a refinement of , denoted 1 1, if R1 R .

partial or#er relation and poset

reflexive, antisymmetric and transitive

Graph: Basic Terminology # #irecte# $rap% & $ (', () consists of a finite set ' and a binary relation on '. #n un#irecte# $rap% & $ (', () consists of a finite set ' and a set of multisets of t!o elements from '.

)#* : % (%)

(v, +) ( if and only if + )#*(v) ,(v) = -v. / )#*(v): the nei$%bor%oo# of v

&01 = (', (01): the reversal of &, !here

(01 $ -(x, y)1 (y, x) (.

G = (V , E ),

E = {(x,y) V V|x y and (x,y) E}:

the

complement of &

# &raph is complete if every pair of distinct vertices is

adjacent.

& = (', () isomorp%ic to &'=('', S')

if a bijection f: ''' satisfyin&, for all x, y ', (x, y) ( (f(x), f(y)) ('

 'ub&raphs
# (partial) sub$rap% 2 of a &raph &=(', () if 2=('', (') satisfyin& '' ' and (' ( in#uce# sub&raph: (iven a subset ) ' of the vertices, &)=(), ()), the sub&raph in#uce# by ) if ()$)xy (1x ) and y )*. spannin$ sub&raph: +he sub&raph contains all vertices of &. span: # subset S ( of the ed&es spans the sub&raph 2=('s, S), !here 's $)v '1v is an endpoint of some ed&e of S*.

clique

# subset ) ' of r vertices is an r0cli"ue if it induces a complete sub&raph, i.e., if &) 3r4 maximal cli"ue maximum cli"ue

5e$ree for di&raphs

out0#e$ree : #/(x) =1)#*(x)1 in0#e$ree : #0(x) =1-y '1x )#*(y).1
d + ( x) = d ( x) =| E| x V x V

sin6 vertex, isolate# vertex

5e$ree for (undirected) &raphs #(x) $ the si,e of its adjacent set,
1

# ( x ) =
x '

Paths and Circuits

c%ain : # sequence of vertices [v0, v1, v2 ,...vl ] is a chain of len&th l in & if v i 1v i E or v i v i 1 E for i$1, , ...,l. pat% : # sequence of vertices [v0 , v1, v2 ,...vl ] is a pat% from v7 to vl of len$t% l in & if v i 1v i E for i$1, , ...,l. simple path: .t does not include the same ed&e t!ice. elementary path(or chain): # path or chain in & is called elementary if no vertex occurs more than once. connecte# $rap% : # &raph & is connecte# if bet!een any t!o vertices there exists a path in & joinin& them. stron&ly connecte# $rap% : # &raph & is stron$ly connecte# if for any t!o vertices x and y there exists a path in & from x to y. elementary cycle(circuit) : # cycle [v0 , v1, v2 ,...vl , v0 ] is a elementary cycle if vi v* for i*. c%or#less cycle : # simple cycle [v0 , v1, v2 ,...vl , v0 ] is c%or#less if vi v*( for i and * differin& by more than 1 mod l/1. Theorem 5.1. .n a (directed or undirected) &raph !ith n vertices, if there is a path from vertex v1 to vertex v , then there is a path of no more than n01 ed&es from v1 to vertex v . bipartite $rap% : #n undirected &raph &=(',() is bipartite if its vertices can be partitioned into t!o disjoint stable sets '=S1/S . complete bipartite $rap% : # bipartite &raph &=(S1,S ,() is complete if for every xS1 and yS !e have xy(, i.e., every possible ed&e that could exist does exist.

Multigraphs and Weighted Graphs

multi$rap% : Let & = (',(), !here ' is a set and ( is a multiset of ordered pairs from ' '. & is so called a directed multi$rap%. +ei$%te# $rap% : & = (',(,f,$), or & = (',(,f), or & = (',(,$), !here ' is a set of vertices, ( is the set of ed&es, f is a function !hose domain is %, and $ is a function !hose domain is (. +he function f is an assi&nment of !ei&hts to the vertices, and the function $ is an assi&nment of !ei&hts to the ed&es. +he !ei&hts can be numbers, symbols, or !hatever quantities.

ome notations
3n : t%e complete $rap% on n vertices or n0cli"ue4 8n : t%e c%or#less cycle on n vertices or n0cycle4 9n : t%e c%or#less pat% $rap% on n vertices or n0pat%4 3m,n : t%e complete bipartite $rap% on m/n vertices partitione# into an m0stable set an# an n0stable set4 31,n : t%e star $rap% on n/1 vertices4 m3n : m #is*oint copies of 3n4 (xample: C4 = 2K2 and Kn,n = 2kn

hortest Paths in Weighted Graphs

len$t% : +he len&th of a path in & is defined to be the sum of the len&ths of the ed&es in the path. s%ortest pat% problem : determine a shortest path from one vertex to another vertex in &. 3ij4stra S%ortest 9at% )l$orit%m : a z 1. .nitially, let 9 = -a. and : = ' 0 -a.. for every vertex t in :, let index l(t) = +(a, t). (l(t) $ if (a, t) (.) l(t), in#ex of t +it% respect to 9: denotes the len&th of a shortest path amon& all paths from a to t that do not include any other vertex in :. . 'elect the vertex in : that has the smallest index !ith respect to 9. Let x denote this vertex. -. .f x is z, stop. .f not, let 9; = 9 -x. and :; = : 0 -x.. 5or every vertex t in :;, compute its index !. r. t. 9; by l;(t) = min <l(t), l(x) / +(x, t)= 1. 6epeat 'teps and - usin& 9; as 9 and :; as :.

 Theorem. 5i*6stra;s s%ortest pat% al$orit%m correctly fin#s t%e len$t% of a s%ortest pat% from a to z4 Proo!. (8rove by induction) Basis Step (i $ 1). obvious Inductive Step. #ssume that for all 6 > i, the 6th time !e arrive at 'tep , l(t) is the len&th of a shortest path from a to t. 9?f t%ere is a pat% from a to a vertex + +%ose len$t% is less t%an l(t), t%en + is not in :4: #ssume that + :: Let ) be a shortest path from a to +, let x be the vertex nearest a on # that is in +, and let u be the predecessor of x on #. +hen u is not in :, so u !as chosen at 'tep durin& a previous iteration. By induction hypothesis, l(u) is the len&th of a shortest path from a to u. l(x) l(u) / +(u, x) len&th of # ; l(t). +his inequality sho!s that v is not the vertex in : !ith minimum l(t), contradicts to the assumption. .t sho!s that if there !ere a path from a to + !hose len&th is less than l(t), + !ould already have been selected at 'tep and remove from +. +herefore, every path from a to t of len&th at least l(t). <. =. 3. 9=ach time a vertex t is inserted into 9, !e have l(t) $ (a, t), the len&th of shortest path from a to t.: 98 forms a shortest0path tree.:

>

"ulerian Paths and Circuits

L. =uler, the father of the &raph theory solved the ?ni&sber&@s brid&e problem, 1>-7 eulerian pat% problem : a path that traverses each ed&e in the &raph once and only once. Theorem 5.#. )n un#irecte# $rap% possess an eulerian pat% if an# only if it is connecte# an# %as eit%er zero or t+o vertices of o## #e$ree4 Proo!. () 'uppose that the &raph possess an eulerian path. .t must be connected. "hen the eulerian path is traced, !e observe that every time the path meets a vertex, it &oes throu&h t!o ed&es !hich are incident !ith the vertex and have not been traced before. +hus, except for the t!o vertices at the ends of the path, the de&ree of any vertex in the &raph must be even.

() omitted.
<. =. 3. Theorem 5.$. )n #irecte# $rap% possess an eulerian circuit if an# only if it is connecte# an# t%e incomin$ #e$ree of every vertex is e"ual to its out$oin$ #e$ree4 )n #irecte# $rap% possess an eulerian pat% if an# only if it is connecte# an# t%e incomin$ #e$ree of every vertex is e"ual to its out$oin$ #e$ree +it% t%e possible exception of t+o vertices4 @or t%ese t+o vertices, t%e incomin$ #e$ree of one is one lar$er t%an its out$oin$ #e$ree, an# t%e incomin$ #e$ree of t%e ot%er is one less t%an its out$oin$ #e$ree4

%amiltonian Paths and Circuits

%amiltonian pat% problem : a path that passes throu&h each of the vertices in a &raph exactly once. Bo simple necessary and sufficient condition is 4no!n. Theorem 5.&. Aet & be a linear $rap% of n vertices4 ?f t%e sum of t%e #e$rees for eac% pair of vertices in & is n 0 1 or lar$er, t%en t%ere exists a %amiltonian pat% in &4 Proo!. (1) & is connected: 'uppose & has t!o or more disconnected components. Let v1 be a vertex in one component that has n1 vertices and v be a vertex in another component that has n vertices. 'ince the de&ree of v1 is at most n1 0 1 and the de&ree of v is at most n 01, the sum of their de&rees is at most n1 / n 0 > n 0 1, contradicts to the assumption. ( ) Construct a hamiltonian path: Let there be a length p-1 (p < n) path, (v1, v , vB, !, vp)4 Both v1 an# vp are adjacent only to the vertices that are in the path4

 :%ere is a cycle containin$ exactly t%e vertices v1, v , vB, !, vp4 #ssume v1 is adjacent to v i > p4
1

,v i2 , ...,v ik ,

!here 1 > i* then !e

k

.f vp is adjacent to one of v i 1,v i 1, have the cycle.

1 2

...,v ik 1,

.f vp is not adjacent to any one of v i 1,v i 1, ...,v i then vp is adjacent to at most p0601 vertices. Contradicts to the assumption.
1 2

1,

8ic4 a vertex vx that is not in the cycle. Because & is connected, there is a vertex v6 that is not in the cycle !ith an ed&e bet!een vx and v6 for some v6 in )v1, v , vB, !, vp*. "e no! ha'e the path (vx, vk, vk+1, , vj-1, vp, vp-1, ,vj, v1, v2, v3, , vk-1), *hich contains p edges. 6epeat the fore&oin& construction until !e have a path !ith n 0 1 ed&es. <. =. 3.

1E

 Theorem 5.5. :%ere is al+ays a %amiltonian pat% in a #irecte# complete $rap%4 Proo!. Let there be a len&th p01 (p > n) path, (v1, v , vB, !, vp)4 Aet vx be a vertex that is not included in this path, and there is no ed&e from vx to v1. Fo!ever, (v1, vx) &4 'uppose that (vx, v ) is also an ed&e in the path. 6eplace the ed&e (v1, v ) in the ori&inal path !ith the t!o ed&es (v1, vx) an# (vx, v ) so that the vertex vx !ill be included in the ar&ument path. .f there is no ed&e from vx to v , then there must be an ed&e (v , vx) in the path and !e can repeat the ar&ument. .f !e find that it is not possible to include vertex v6 in any au&ment path by replacin& an ed&e (v6, v6/1) in the ori&inal path !ith t!o ed&es (v6, vx) and (vx, v6/1)!ith 1 6 p01, then !e conclude that there must be an ed&e (vp, vx) in the &raph. "e can repeat the ar&ument until all vertices in the &raph are included in the ar&umented path. <. =. 3. +here is no &eneral method of solution to the problem of provin& the non0existence of a hamiltonian path or circuit in a &raph.

11

The Tra'eling alesperson Pro+lem

:ravelin$ salesperson problem : Let & = (', (, +) be a complete &raph of n vertices, !here + is a function from ( to the set of positive real numbers such that for any three vertices, i, *, 6 in ' +(i, *) / +(*, 6) +(i, 6) +he travelin& salesperson problem as4s for a hamiltonian circuit of minimum len&th. (,P-complete pro+lem) ,earest nei$%bor method 1. 'tart !ith an arbitrarily chosen vertex, and find the vertex that is closest to the startin& vertex to form an initial path of one ed&e. . Let x denote the latest vertex that !as added to the path. #mon& all vertices that are not in the path, pic4 and one that is closest to x, and add to the path the ed&e connectin& x and this vertex. 6epeat this step until all vertices in & are included in the path. -. 5orm a cycle by addin& the ed&e connectin& the startin& vertex and the last vertex added. Theorem 5... 5or a &raph !ith n vertices, let # be the total distance of a hamiltonian circuit obtained accordin& to the nearest0nei&hbor method and #7 be the total distance of a minimum hamiltonian circuit. +hen lgn + .
d d 0 1 2 1 2

/actors o! a Graph
60factor of a &raph: a spannin& sub&raph !ith the de&ree of each of its vertices bein& 6. # hamiltonian circuit of a &raph is 0factor.

1-

Planar Graphs
planar &raph : # &raph is said to be planar if it can be dra!n on a plane is such a !ay that no ed&es cross one another, except, of course, at common vertices. re$ion : # re&ion of a planar &raph is defined to be an area of the plane that is bounded be ed&es and is not further divided into subareas. # re&ion is said to be finite if this area is finite, and is said to be infinite if its area is infinite. Clearly, a planar &raph has exactly one infinite re&ion. Theorem 5.0. @or a connecte# planar $rap%, v - e + r = ((uler;s formula) !here v, e, and r are the number of vertices, ed&es, and re&ions of the &raph, respectively. #pplication of =uler@s formula : ?n any connecte# planar $rap% t%at %as no loops an# %as t+o or more e#$es, e 3v -6. Theorem 5.1. (2urato*s3i) ) $rap% is planar if an# only if it #oes not contain any sub$rap% t%at is isometric to +it%in vertices of #e$ree to eit%er 3C or DB,B4

11