Anda di halaman 1dari 23

Jerick Mari A.

Tancinco November 10, 2011 RELATIONSHIP OF GRAPHS AND GREEDOIDS INTRODUCTION In 1935, Whitney introduced the notion of matroids in studying planar graphs. He dened concepts like independence and rank in graphs and these were found to be connected also to linear independence and dimension in linear algebra. Eventually, Rado and Edmonds discovered its relation to greedy algorithms and the problems that these algorithms can solve. They serve as the mathematical structure behind the greedy algorithms. Their existence is signicant in such a way that they can determine the optimality of a given greedy strategy on a problem. However, their non-existence might not mean the strategy is not optimal In 1982, Korte and Lovasz had generalized matroids to greedoids in by relaxing the denition of matroids. With this, more greedy algorithms were then covered by later studies. To dene, a greedoid is a pair (C, F ), where C is a nite set called the ground set and F is a non-empty collection of subsets of C called the family of feasible sets, that satises the following properties: (accessibility) for every non-empty feasible set X there exists x X such that X\{x} is also feasible; and, (augmentation) for any feasible sets X and Y such that |X| = |Y | + 1, there exists x X\Y such that Y {x} is also feasible. Meanwhile, a matroid is a greedoid with a stricter accessibility property, that is, (hereditary) if X Y and Y is a feasible set, then X is also feasible. In graphs, greedoids can be obtained in dierent ways and vice versa. The relationship of the two has created plenty of studies, open problems, and applications.

DEFINITIONS The Greedy Algorithm. Let (C, F ) be a set system with an associated function W : F R. Let X be an empty set. Choose an x C\X such that 1. X {x} is feasible 2. W (X {x}) W (X {y}) for all y such that X {y} is feasible. Set X = X {x} and repeat until X can no longer be augmented. The Autonomous Greedy Algorithm. Let (C, F ) be a set system with an associated function f : C 2C R+ . Let X be an empty set. Choose an x C\X such that f (x, X {x}) f (y, X {y}) for all y. Set X = X {x} and repeat until X = C. The dierence between the two algorithms is that the latter picks an element regardless of whether it yields a feasible set or not. Graph Denitions A graph is a pair (V, E) where V is a nite non-empty set whose elements are called vertices while E is a list of unordered pairs of vertices whose elements are called edges. If the pairs of vertices are ordered, then (V, E) is called a digraph. Let G be a graph. Then, the girth of G is the length of the shortest cycle contained in G. Isolated vertices in G are vertices of degree 0. Graph G is said to be n-connected if G has no set of n 1 vertices whose deletion results to a disconnected graph. A matching M of G is said to be uniquely restricted if the induced subgraph of the vertices saturated by M has only one perfect matching, that is, M itself. Let (G) denote the cardinality of a maximum matching in G. Let G be a graph on m vertices and let {Hi : 1 i m} be a family of graphs. Then, the corona of G and {Hi : 1 i m}, denoted by G {Hi : 1 i m}, is the graph obtained by joining each vertex vi of G to all the vertices of Hi . If H1 = H2 = . . . = Hn = H, then the corona is written as G H, called the corona of G and H. The neighborhood of v, denoted by N (v), where v is a vertex of a graph G is dened as N (v) = {w : vw is an edge in G} . (1) 2

The neighborhood N (A), where A is a subset of V , is dened as N (A) = {v : v is in V \A, and N (v) A = } . (2)

Then, A N (A) is the closed neighborhood of A, denoted by N [A]. A stable set of a graph G is a set of pairwise non-adjacent vertices in G. A stable set of maximum size is a maximum stable set, and the cardinality of a maximum stable set is denoted by (G), called the stability number of G. Thus, a local maximum stable set A is a maximum stable set in the subgraph induced by N [A]. Let the family of local maximum stable sets of a graph G be dennoted by (G). In a graph, if (G) + (G) = |V (G)|, then it is called a Knig-Egervry o a graph. A graph G is well-covered if every maximal stable set of G is also a maximum stable set. If, in addition, G has no isolated vertices and |V (G)| = 2(G), then G is very well-covered. Let (C, F ) be a set system with an associated weight function w : C + R . The value function v : C 2C R+ is dened as w(y) : x X F, X A , if such X exists min yX v(x, A) = 1 + w(y) , otherwise.
yC

(3) In other words, v(x, A) is the minimum weight of a feasible set that is contained in A that must also contain x. The shortest path function W : 2C R+ is dened as v(x, A). (4) W (A) =
xA

The maximum path function W : 2C R+ is dened as W (A) = max v(x, A).


xA

(5)

The Shortest Path Problem. Given a set system (C, F ) and the shortest path function W , nd Xk F with |Xk | = k such that W (Xk ) = min {W (Y ) : Y F, |Y | = k} . (6)

The Maximum Path Problem. Given a set system (C, F ) and the maximum path function W , nd Xk F with |Xk | = k such that W (Xk ) = min {W (Y ) : Y F, |Y | = k} . 3 (7)

The Ordered Path Problem. Given a set system (C, F ) and the value function v, nd an x1 . . . x|C| of C such that v(xi , C) v(xj , C) for i j. Greedoid Denitions The rank of a greedoid D = (C, F ) is the maximum cardinality of its feasible sets, and it is denoted by r(C). Let A be a subset of C. The rank of A is dened as r(A) = max {|X| : X A, X is f easible} . Then, the rank closure of A is dened as (A) = {x C : r(A {x}) = r(A)} . Moreover, the nullity of A in D is dened as nulD (A) = |A| r(A). (10) (9) (8)

A greedoid is an antimatroid if, for every feasible sets X and Y , X Y is also feasible. A greedoid is said to satisfy the intersection property if, for every feasible sets X and Y , X Y is also feasible. A greedoid has the local union property if, for every feasible sets X, Y , and Z with X and Y subsets of Z, X Y is a feasible set. If, instead, X Y is feasible, then it satises the local intersection property. A set system (E, F ) satises the union closure property if, for every feasible sets X and Y , (X) (Y ) (X Y ). Interval greedoids are greedoids satisfying the local union property while local poset greedoids are greedoids satisfying both the local union and the local intersection properties. The lift of a greedoid D = (C, F ) is the greedoid L = (C, F (L)) where F (L) = {A C : nulG (A) 1} . (11)

The greedoid D is n-connected if, for any feasible subsets X, there exists a set Y with |Y | = min {n, r(C) |X|} such that for all Y Y , X Y is feasible. The greedoid polynomial fG (t, z) is dened as fG (t, z) =
AC

tr(C)r(A) z |A|r(A) .

(12)

REVIEW OF LITERATURE Greedoids Arising from Graphs In Klappenckers paper (2008), two types of greedoids arising from graphs were presented. Theorem 1: Let G = (V, E) be an undirected graph. Then, the graphic matroid of G is the set system (E, S) where S is the set of edge sets of any forest in G. Theorem 2: Let G = (V, E, r) be a (directed) graph with a root vertex r. Then, the (directed) branching greedoid of G is the set system (E, S) where S is the set of the edge sets of rooted trees (rooted arborescences) in G. Johnson (2009) presented a detailed development of matroids in his paper. It included a variety of matroids obtainable from graphs. Graphic matroids, also known as circuit or cycle matroids, were also covered. Other classes of matroids are presented below. Theorem 3: In a graph G, the cographic matroid of G is the set system D such that circuits in D are the cut-sets of G. If G is a connected planar graph, then the cographic matroid of G is the graphic matroid of its dual graph. Theorem 4: The bicircular matroid of a graph G is the set system (E, F ) where F is the set of edge sets of all subgraphs of G having at most one cycle in every component of those subgraphs. Theorem 5: Consider a bipartite graph G with bipartition {X, Y }. Then, a transversal matroid of G is a set system (X, F ) where F is the collection of all subsets of X such that for each I F there exists a matching in G that saturates the vertices in I. Theorem 6: In an undirected graph G = (V, E), a matching matroid in G is a set system (J, F ) where J V and F is the collection of all subsets of J such that for each I F there exists a matching in G that saturates the vertices in I. If G is a bipartite graph and J is one of the bipartition, then (J, F ) is a transversal matroid. 5

Theorem 7: Consider two sets of vertices S and T in a directed graph D, not necessarily distinct. Then, the set system (S, F ) is a gammoid where F is the collection of subsets of S such that there exist pairwise vertex-disjoint directed paths from every vertex in I F to a vertex in T . If S and T partitions V (D) and every edge in G goes from S to T , then (S, F ) will be a transversal matroid. In fact, gammoids are the closure of transversal matroids under the operations of minors and duality. Some of the developments in matroids concern about operations on matroids and their graph theoretic interpretations. The following theorems were proved. Theorem 8: The graphic and cographic matroids of a graph G are dual to each other. If the cographic matroid of G is also graphic, then G is planar. Such matroids are called planar matroids. Theorem 9: The dual of a transversal matroid or a gammoid is always a gammoid. Theorem 10: Let M be the graphic matroid of a graph G witha feasible and let S and T be disjoint sets of edges of G. Then, i) the contraction of M by a, denoted by M/a, is the graphic matroid of the graph obtained by contracting edge a from G. ii) the deletion of M by a, denoted by M a, is the graphic matroid of the graph obtained by deleting edge a from G. iii) the minor M \S/T of M is the graphic matroid of the graph obtained by deleting the edges in S and contracting the edges in T from G. Theorem 11: If a graph G is biconnected, then it cannot be written as a non-trivial sum of matroids. Simple matroids were also a part of development of matroids. A theorem was proven relating such matroids to graphs. Theorem 12: The graphic matroids of simple graphs are simple. Okamoto and Nakamura (2003) presented two antimatroids arising from rooted digraphs. 6

Theorem 13: Let G = (V, E, r) be a rooted graph. Then, i) (V, F ) is the point-search antimatroid of G where F is the set of subsets I V such that, for every v I, there exists a (r, v)-path in the subgraph induced by I {r}. ii) (E, F ) is the line-search antimatroid of G where F is the set of subsets J E such that, for every e J, there exists a path from r containing e in the subgraph induced by J. They showed the following result relating antimatroids to rooted digraphs. Theorem 14: Given an antimatroid F , i) F is the point-search antimatroid of a rooted digraph if and only if F does not contain a minor isomorphic to D5 = {, {x}, {y}, {x, y}, {x, y, z}}. ii) F is the line-search antimatroid of a rooted digraph if and only if F has no minor isomorphic to D5 or the point-search antimatroids of A, B, Cm,n , Dl,m,n (l, m, n 1) where

According to Tedford (2007), a generalization of the undirected branching greedoid G of a rooted graph G can be found by taking its lifts. Denition 1: The lift of an undirected branching greedoid D = (C, F ) is the greedoid L = (C, F (L)) where F (L) = {A C : nulG (A) 1} . The k th -lift Lk (G ) can be obtained by recursively applying this procedure. Some properties of the original greedoid were found to be passed down from one lift to another: Theorem 15: If n 2 and Lk (G ) is n-connected, then Lk (G ) is n-connected for all i k. 7

Let e be a bridge and v a cutvertex of a rooted graph G. Let Cp (v) be the component of G v containing the root p and let Cp (e) be the component of G e containing p. Then, the following theorem was proved. Theorem 16: If Lk (G ) is 2-connected, then i) |E(Cp (v))| < |V (Cp (v))| + k 1 and ii) |E(Cp (v))| > |E(G)| k 2. It was proven in his paper that the converse is not true.

Eaton (2008) considered T = (V, E), a multi-rooted directed tree, and s V , a sink that is not a leaf. Since T is a multi-rooted, s exists and T s is disconnected. Denote by C1 , C2 , . . . , Ck be the components of T s, and dene Vi = V (Ci ) {s}. Let Ti be the graph induced by Vi . It was easily seen that E = E1 E2 . . . Ek and Vi Vj = {s} for all i = j. From the multi-rooted directed tree T , a greedoid G(T ) = (E, F ) was considered where F is the set of edge sets of any acyclic union of rooted arborescences in T . Then, the following theorem was proved. Theorem 17: Given a multi-rooted directed tree T , G(T ) = k G(Ti ). i=1 Levit and Mandrescu formed greedoids out of various graphs called the local maximum stable set greedoids (G). The ground sets of these greedoids are the vertex sets of the graphs. The following theorems are dedicated in nding these greedoids. Theorem 18 (Levit and Mandrescu): i) For every forest F , (F ) is a greedoid (2002). ii) Let G be a well-covered graph of girth at least 6 which has no component isomorphic to C7 . Then, (G) is a greedoid (2008). iii) If (G) is a greedoid, then (Ck ) (G) = for every cycle Ck , k 4, induced by the sets in (G) (2007,2009). iv) If G is a graph with a unique cycle C3 , then (G) is a greedoid (2009). 8

Some of their discoveries were both necessary and sucient in forming greedoids. Theorem 19 (Levit and Mandrescu): i) For a bipartite graph G, (G) is a greedoid if and only if all its maximum matchings are uniquely restricted (2004). ii) If G is a triangle-free graph, then (G) is a greedoid if and only if all of the maximum matchings of G are uniquely restricted and the closed neighborhood of every local maximum stable set of G induces a Knig-Egervry graph (2007). o a iii) Let Ck be the unique cycle of the graph G. Then, a) if k = 2q 4, then (G) is a greedoid if and only if all maximum matchings of G are uniquely restricted. b) if k = 2q + 1 5, then (G) is a greedoid if and only if the closed neighborhood of every local maximum stable set of G induces a Knig-Egervry graph. (2009) o a iv) Let G be a very well-covered graph. Then, (G) forms a greedoid if and only if G has a unique maximum matching (2010a,2011). Greedoids also exist under the inuence of graph operations. Theorem 20 (Levit and Mandrescu): i) The family (H K1 ) is a greedoid (2008a). ii) Let G be the disjoint union of the family of graphs G1 ,G2 ,. . .,Gn where n 2. Then, (G) is a greedoid if and only if (G1 ),(G2 ),. . .,(Gn ) are greedoids (2010b). iii) Let Z be the Zykov sum of G1 ,G2 ,. . .,Gn such that (Z) > 1. Then, (Z) is a greedoid if and only if the following are satised: a) (G1 ),(G2 ),. . .,(Gn ) are greedoids b) there is a unique 1 k n such that Gk is not complete c) (Z) = (Gk ). (2010b) 9

iv) Let G = X {H1 , H2 , . . . , Hn } where H1 ,H2 ,. . .,Hn are non-empty graphs and X is any graph. Then, (G) is a greedoid if and only if (H1 ),(H2 ),. . .,(Hn ) are greedoids (2010b). In a very well-covered graph, several equivalent statements were proved (Levit and Mandrescu, 2011). Theorem 21: Let G be a very well-covered graph. Then, the following are equivalent statements: i) (G) is a greedoid. ii) G has a uniquely restricted maximum matching. iii) G has an alternating cycle-free maximum matching. iv) G has an alternating C4 -free maximum matching. v) Every maximum matching in G is alternating cycle-free. vi) Every maximum matching in G is alternating C4 -free. vii) All of the maximum matchings of G are uniquely restricted. For the following theorems, Levit and Mandrescu (2008b) not only searched for more greedoids out of graphs but also indicated the class of greedoids where they belong. Theorem 22: The family (G) of a graph G satises the accessibility property if and only if (G) forms an interval greedoid. Theorem 23: For any graph G, (G) is an antimatroid if and only if the following are satised: i) (G) is accessible ii) G is a unique maximum stable set graph. Corollary 1: For any tree T , the following are equivalent statements: i) (T ) is an antimatroid ii) T is a unique maximum stable set graph 10

iii) T has a maximum stable set S such that |N (v) S| 2 for every v V (T )\S. Theorem 24: For any graph G, the following are equivalent: i) (G) is a matroid ii) S simp(G) for every maximum stable set S of G iii) G is a simplicial graph and every non-simplicial vertex belongs to at least two dierent simplices. Corollary 2: If G is a triangle-free graph, then the following are rquivalent: i) (G) is a matroid ii) S pend(G) isol(G) for every maximum stable set S of G iii) G has K1 , K2 and graphs having unique maximum stable sets, namely, sets of their pendant vertices. Corollary 3: If T is a tree of order at least 3, then the following are equivalent: i) (T ) is a matroid ii) (T ) is a trimmed matroid iii) pend(T ) is a unique maximum stable set of T . Theorem 25: If S simp(G) for every maximum stable set S of G, then (G) is a local poset greedoid.

Graphs Arising from Greedoids Acketa (1993) converted rank 2 non-trivial greedoids to graphs with marked vertices using the denition below: Denition 2: i) Each edge has at least one marked vertex.

11

ii) Each marked vertex is connected by an edge to at least one vertex of each non-incident edge. In other words, feasible sets of size 2 become the edges and those of size 1 as the marked vertices. Then, all the rank 2 greedoids on at most 7 elements were generated. The data gathered were presented in the table below: N GG(n) GD(n) GC(n) 2 1 2 3 3 3 8 12 4 9 25 43 5 25 70 126 6 80 200 380 7 265 592 1114

where GG(n) = the number of non-isomorphic greedoidic graphs on n vertices, GD(n) = the number of non-isomorphic rank 2 greedoids on n elements, and GC(n) = the number of greedoids on n elements associated to all non-isomorphicgreedoidic graphs on n vertices. As n increases, it was observed that GG(n), GD(n), and GC(n) approximately triples, and the ratios GD(n)\GG(n) and GC(n)\GD(n) were considerably stable. Moreover, the trees adjoined to full rank 2 greedoids were found to be stars and double stars only. The number of non-isomorphic rank 2 greedoids associated with certain graphs were also determined: Theorem 26: 1 + (1)n 1 2 n 2n 9 + i) For trees, there were 4 2 isomorphic rank 2 greedoids; ii) for complete graphs Kn , there were two; and, iii) for Kn with one edge deleted, there were four. Broersma and Li (1992) made discoveries on the basis graphs of branching greedoids. They proved the following theorems. 12 associated non-

Theorem 27: Let D be a 2-connected branching greedoid of a rooted multigraph D = (V, E, r). Then, B(D ) is connected. Theorem 28: Let D be a 3-connected branching greedoid of a rooted multigraph D = (V, E, r). Then, i) B(D ) is ( 1)-connected ii) If v V \ {r}, Bv = {X | X is a basis of D with leaf v} induces a ( 1)-connected subgraph of B(D ). Theorem 29: Let D = (V, E, r) be 3-connected and let X be a basis of D . Then, for any v V \{r}, D has 1 bases X1 , X2 , . . . , X1 that contains v as a leaf and X {X1 , X2 , . . . , X1 } is a fan in B(D ). Broersma and Li showed that in a multigraph D whose root r is a cutvertex, the basis graph of D is isomorphic to the Cartesian product graph of the basis graphs of D1 , D2 , . . . , Dn , where D1 , D2 , . . . , Dn are the components of D r. Also, in a rooted multigraph H that has a vertex v with indegree 1 and d(r, v) = 1, the basis graph of D is isomorphic to the basis graph of (D/v) .

Li, Lara and Campo (2005) considered a rooted graph (V, E, r) and let IG,k be the set of edge sets of all connected rooted subgraphs of G such that, for X IG,k , |X| k and, if |X| = k, then the induced subgraph of X must be a spanning subgraph of G. Then, they proved the following theorem. Theorem 30: (E, IG,k ) is a greedoid where k |V | 1. Moreover, the basis graph of (E, IG,k ) is a spanning subgraph of the LEE-graph of G. Then, the previous theorem was applied in proving the next result. Theorem 31: The LEE-graph of a graph G is connected if G is 2-connected. Polynomials Gordon and McMahon (1989) has dened a greedoid polynomial f (D; t, z) over a rooted digraph Dv where v is its root vertex. The greedoid polynomial was used to relate the rooted digraph to its directed branching greedoid through the following theorems, and vice versa. 13

Theorem 32: Let Dv be a rooted digraph. Then, i) the coecient of tr(Dv )1 is the outdegree of v. ii) f (Dv ; 1, 0) is the number of feasible sets. iii) f (Dv ; 0, 1) is the number of spanning sets. iv) f (Dv ; 0, 0) is the number of bases. Theorem 33: The following statements are equivalent in a rooted digraph D. i) D is an arborescence. ii) f (D; 0, 1) = 1. iii) In every term, the exponent of z is less than or equal to the exponent of t. iv) There is no pure z term. Theorem 34: Let D1 and D2 be rooted arborescences. If f (D1 ) = f (D2 ), then D1 and D2 are isomorphic. McMahon (1993) proved a theorem about the connection between rooted graphs and their associated branching greedoid. Theorem 35: If G1 and G2 are rooted graphs (digraphs) with no greedoid loops and isolated vertices, then their associated branching greedoids are isomorphic if and only if G1 and G2 are isomorphic. This implies that such rooted graphs (digraphs) must be equivalent to their associated branching greedoids. McMahon also found relationships between rooted graphs (digraphs) and the greedoid polynomial. Theorem 36: Let G be the branching greedoid of a rooted graph G. If fG (t, z) = (z + 1)a f1 (t, z) where z + 1 does not divide f1 (t, z), then a is the number of greedoid loops in G.

14

Theorem 37: Let D be the associated greedoid of a rooted digraph D with no greedoid loops. Then, D has a directed cycle if and only if z + 1 divides fD (t, z). Using the characteristic polynomial G , the number of minimum tree covers of a graph G was determined by Gordon and McMahon (2001) through this theorem. Theorem 38: If deg(G ()) = k and ak is the coecient of k , then the number of minimum tree covers of G is (1)r(G)k ak . The characteristic polynomial was also shown to determine the number of acyclic orientations O of a rooted graph. Theorem 39: Let G be a rooted graph with root r and let O(G) be the collection of all acyclic orientations of G which create no greedoid loops in the directed branching greedoid associated with O. Then, |O(G)| = (1)r(G) G (0). Lastly, theorems were also proved when the rooted graph considered is a rooted fan. Theorem 40: Consider the rooted fans Fm and Fn . Then, i) Fm () | Fn () if and only if m | n. ii) Fm ()/ 1 is irreducible if and only if m is prime. McMahon, Shimkus and Wolfson (2003) formed a class of greedoids from chordal graphs G, called the simplicial shelling antimatroids. Theorem 41: Let G be a chordal. Then, A(G) = (V, F ) is the simplicial shelling antimatroid of G such that if I F then the vertices in I can be ordered so that the rst vertex v1 is simplicial in G, the second vertex v2 is simplicial in G v1 , the third vertex is simplicial in G {v1 , v2 } , and so on. In their paper, antimatroids can be dened either in terms of feasible sets or in terms of convex sets. They proved the following theorem in their paper. Theorem 42: In a chordal graph G = (V, E) with components G1 , G2 , . . . , Gm , C V corresponds to a free convex set of A(G) if and only if C V (Gi ) induces a clique in Gi for every component Gi of G. 15

In their next result, they considered a connected chordal graph, its decomposition, and the characteristic polynomial. Then, the following was also proven. Theorem 43: Let G = Kn be a connected chordal graph that has a decomposition R1 S1 R2 S2 . . . Sn1 Rn where Ri and Sj are complete. If A(G () = ( 1)m 1 () where 1 1 (), then m is the size of the smallest Sj where pasting occurs. Lastly, it was shown that the number of blocks b(G) in a chordal graph G can be determined using the derivative of a characteristic polynomial. Theorem 44: Let G be a chordal graph. Then, b(G) = 1+(1)|A(G)| A(G) (1). In Gordons paper (2008), his denition of the Tutte polynomial f (G; x, y) was in terms of graphs. However, it can be dened equivalently in terms of greedoids, specically, the graphic matroids of those graphs. Thus, even the results in their paper has an underlying greedoid being used. Theorem 45: Let G be a graph. Then, i) the number of subsets of edges of G is f (G; 2, 2) ii) the number of spanning trees of G is f (G; 1, 1) iii) for the number of subsets of edges of G that contain a spanning tree, f (G; 1, 2) iv) for the number of subsets of the edges of G that are contained in a spanning tree, f (G; 2, 1) v) for the acyclic orientations of the edges of G, f (G; 2, 0) vi) for the acyclic orientations of the edges of G in which a specied vertex v is the unique source, f (G; 1, 0) vii) for the orientations of G in which every edge is in some cycle, f (G; 0, 2) viii) for the distinct score vectors that arise from orientations of the edges of G, f (G; 2, 1)

16

The Tutte polynomial was also proven to be related with the chromatic , ow and reliability R polynomials. Theorem 46: Let G be a graph with m vertices, n edges and c components. Then, i) (G; ) = c (1)mc f (G; 1 , 0) ii) (G; ) = (1)nm+c f (G; 0, 1 ) iii) R(G; p) = (1 p)nm+c pmc f (G; 0, 1 ) 1p

The Tutte polynomial was also dened over a rooted graph Gv , where v V (G) is its root. Likewise, this polynomial can be equivalently dened over the branching greedoids of rooted graphs. The following theorems on rooted graphs were proven. Theorem 47: Let Gv be a rooted graph. Then, i) the number of subsets of edges of Gv is f (Gv ; 2, 2) ii) for the spanning trees of Gv , f (Gv ; 1, 1) iii) for the subsets of edges of Gv that contain a spanning tree, f (Gv ; 1, 2) iv) for the rooted subtrees of Gv , f (Gv ; 2, 1) v) for the acyclic orientations of the edges of Gv in which v is the unique source, f (Gv ; 1, 0) Theorem 48: Let T1 and T2 be rooted trees. Then, f (T1 ; x, y) = f (T2 ; x, y) if and only if T1 and T2 are isomorphic. For an unrooted tree T , the Tutte polynomial can be dened over the pruning greedoid of T . However, the conjecture below analogous to the previous theorem was shown to be false. Conjecture 1: Let T1 and T2 be unrooted trees. Then, g(T1 ; x, y) = g(T2 ; x, y) if and only if T1 and T2 are isomorphic.

17

Gordon and McMahon (1997) constructed a way to form a binary tree T (G) from a greedoid G = (C, F ), called as a computation tree. Following the steps below recursively, such graph was formed. Denition 3: i) If G has no feasible sets of size 1, then T (G) is a trivial rooted tree whose only vertex is labeled by G. ii) If {e} is feasible in G, then the root of T (G) is labeled by G. The left and right subtrees of T (G) are T (G/e) and T (G e), respectively. Let G1 , G2 ,. . ., Gk be the leaves of the computation tree of G. Then, Fi was dened as the set of elements of C that were contracted along the (G, Gi )path in T (G), and the external activity extT (Fi ) of Fi was dened as those that were neither deleted nor contracted in the process. Then, they proved the following theorem. Theorem 49: Let G = (C, F ) be a greedoid and T (G) be any of its computation tree with kT leaves, then i) F = Fi | 1 i k and ext(Fi ) (Fi )\Fi for all 1 i kT . ii) G is an antimatroid if and only if ext(Fi ) = (Fi )\Fi for all 1 i k. iii) the intervals of the form [Fi , Fi extT (Fi )](1 i k) partition 2C . Furthermore, this partition is unique if and only if G is an antimatroid. In this case, Fi ext(Fi ) = (Fi ). The internal activity intT (B) was dened for bases B of G. An element e C is internally active if it is a feasible coloop in some minor in the path from the root to the leaf corresponding the base B. A computation tree can be made with respect to a total order of C by nding a feasible element in C based on the ordering. Also, denote by i (B) and e (B) the internally and externally active elements of B, respectively, with respect to the total order of C. Theorem 50: Let G = (C, F ) be a matroid with a total order of C. Then, there is a computation tree T (G) such that i (B) = intT (B) and e (B) = extT (B) for all bases B of G.

18

Theorem 51: Let G = (C, F ) be a greedoid with computation tree T (G)and bases B1 , B2 , . . . , Bm . Then, the intervals of the form [Bi intT (Bi ), Bi extT (Bi )](1 i m) partition 2C . Theorem 52: Let be a total order of the ground set of a greedoid G and let T (G) be a computation tree of G. Then, e (B) = extT (B) for all bases B of G. Lastly, an application of Gordon and McMahons results was the proof of the next theorem relating directed graphs to their branching greedoid using the Tutte polynomial f . Theorem 53: Let D be the branching greedoid of a rooted digraph D with no greedoid loops. If fD (t, z) = (z + 1)k f1 (t, z) where z + 1 f1 (t, z), then k is the minimum number od edges needed to be removed from D to have a spanning acyclic rooted digraph.

Graph-theoretic Problems Mielikinen and Ukkonen (2006) mentioned in their paper that the maxia mum matroid-greedoid intersection problem can be reduced to the maximum matching problem on a bipartite graph G = (V, E) with bipartition {X, Y }. Two matroids M1 and M2 were considered. The family of feasible sets of M1 would be the sets I E such that, for each vertex in X, I contains at most one edge incident to that vertex. M2 will be constructed the same way, but with respect to Y . Then, applying a solution to the maximum matroidgreedoid intersection problem also solves the maximum matching problem Moreover, the maximum tree-constrained matching problem on a bipartite graph G = (V, E) with bipartition {X, Y } was also shown as a special case of the maximum matroid-greedoid intersection problem. A matroid N1 and a greedoid N2 were constructed. The family of feasible sets of N1 would be the same as the previously constructed matroids. For N2 , a rooted spanning tree T contained in G was considered. N2 was constructed the same way as the others with an additional condition: the induced subgraphs of the feasible sets must be rooted subtrees of T . Futhermore, the weighted counterparts of these matching problems were also shown as special cases of the maximum weighted matroid-greedoid intersection problem. 19

Boyd (1988) focused on the types of greedoids behind the greedy solution of some path problems in graph theory. Theorem 54: Let (C, F ) be a set system. If the greedy algorithm solves the shortest path problem for an arbitrary weight function w : C R+ , then (C, F ) is a greedoid. Theorem 55: The autonomous greedy algorithm solves the ordered path problem on an accessible set system. Boyd found results involving the minmax path problem. Theorem 56: The greedy algorithm solves the minmax path problem on a set system (C, F ) if any of the following is satised: (i) (C, F ) is an antimatroid. (ii) (C, F ) is a greedoid with the intersection property. (iii) (C, F ) is a local poset greedoid with the union closure property. Theorem 57: For an interval greedoid, if the greedy algorithm solves the minmax path problem, then it also solves the shortest path problem. Due to the close relationship of the minmax path problem with the shortest path problem, Theorems 10 and 11 were used to prove the following results for the shortest path problem. Theorem 58: The greedy algorithm solves the shortest path problem on a set system (C, F ) if any of the following is satised: (i) (C, F ) is an antimatroid. (ii) (C, F ) is an interval greedoid with the intersection property. (iii) (C, F ) is a local poset greedoid with the union closure property.

20

REFERENCES 1. Acketa, D. M. (1993). On rank 2 greedoids [Electronic Version]. Univ. u Novom Sadu Zb. Rad. Prirod.-Mat. Fak. Ser. Mat. 23 (2), 289-303. 2. Boyd, E. A. (1988). A combinatorial abstraction of the shortest path problem and its relationship to greedoids. Retrieved August 19, 2011, from http://www.caam.rice.edu/caam/trs/88/TR88-07.pdf 3. Broersma, H. J. and Li, X. (1992). The connectivity of the basis graph of a branching greedoid [Electronic Version]. Journal of Graph Theory, 16(3), 233-237. 4. Eaton, L. (2008). The characteristic polynomial of multi-rooted directed trees. Retrieved September 10, 2011, from http://dspace.nitle. org/bitstream/handle/10090/5291/Eaton.pdf?sequence=1 5. Gordon, G. (2008). Chromatic and Tutte polynomials for graphs, rooted graphs, and trees [Electronic Version]. Graph Theory Notes of New York LIV, 34-45. 6. Gordon, G. and McMahon, E. (1989). A greedoid polynomial which distinguishes rooted arborescences [Electronic Version]. Proceedings of the American Mathematical Society, 107(2), 287-298. 7. Gordon, G. and McMahon, E. (1997). Interval partitions and activities for the greedoid Tutte polynomial [Electronic Version]. Advances in Applied Mathematics, 18, 33-49. 8. Gordon, G. and McMahon, E. (2001). A characteristic polynomial for rooted graphs and rooted digraphs [Electronic Version]. Discrete Mathematics, 232, 19-33. 9. Johnson, W. (2009). Matroids. Retrieved October 26, 2011, from http://www.math.washington.edu/morrow/ 336 09/papers/Will.pdf 10. Klappenecker, A. (2008). Theory of greedy algorithms. Retrieved October 29, 2011, from https://parasol.tamu.edu/welch/teaching/411. f08/greedy.pdf

21

11. Levit, V. E. and Mandrescu, E. (2002). A new greedoid: The family of local maximum stable sets of a forest. Discrete Applied Mathematics, 124, 91-101. Retrieved August 16, 2011, from http://arxiv.org/PS cache/math/pdf/9912/9912222v1.pdf 12. Levit, V. E. and Mandrescu, E. (2004). Local maximum stable sets in bipartite graphs with uniquely restricted maximum matchings [Electronic Version]. Discrete Applied Mathematics, 132, 163-174. 13. Levit, V. E. and Mandrescu, E. (2007). Triangle-free graphs with uniquely restricted maximum matchings and their corresponding greedoids [Electronic Version]. Discrete Applied Mathematics, 155, 24142425. 14. Levit, V. E. and Mandrescu, E. (2008a). Well-covered graphs and greedoids. Proceedings of the 14th Computing: The Australasian Theory Symposium, Wollongong, NSW, Conferences in Research and Practice in Information Technology Volume, 77, 88-94. Retrived August 16, 2011, from http://crpit.com/confpapers/CRPITV77Levit.pdf 15. Levit, V. E. and Mandrescu, E. (2008b). Interval greedoids and families of local maximum stable sets of graphs. Retrieved September 5, 2011, from http://arxiv.org/PS cache/arxiv/pdf/0811/0811.4089v1.pdf 16. Levit, V. E. and Mandrescu, E. (2009). Greedoids on vertex sets of unicycle graphs. Retrieved September 5, 20ll, from http://arxiv.org/PS cache/arxiv/ pdf/0905/0905.1024v1.pdf 17. Levit, V. E. and Mandrescu, E. (2010a). Very well-covered graphs of girth at least four and local maximum stable set graphs [Electronic Version]. Discrete Mathematics, Algorithms and Applications, 3(2), 245-252. 18. Levit, V. E. and Mandrescu, E. (2010b). Graph operations that are good for greedoids. Discrete Applied Mathematics, 158, 1418-1423. Retrieved September 5, 2011, from http://arxiv.org/PS cache/arxiv/pdf/ 0809/0809.1806v1.pdf 19. Levit, V. E. and Mandrescu, E. (2011). Local maximum stable sets greedoids stemmed from very well-covered graphs. Retrieved August 22

19, 2011, from http://arxiv.org/PS cache/arxiv/pdf/1102/1102. 1142v1.pdf 20. Li, X., Lara, V. N. and Campo, E. R. (2005). Two approaches for the generalization of leaf edge exchange graphs on spanning trees to connected spanning k-edge subgraphs of a graph. Ars Combin., 75, 9pp. Retrieved September 5, 2011, from http://www.combinatorics.cn/ publications/papers/2002/13.pdf 21. McMahon, E. W. (1993). On the greedoid polynomial for rooted graphs and rooted digraphs [Electronic Version]. Journal of Graph Theory, 17, 433-442. 22. McMahon, E. W., Shimkus, B. A. and Wolfson, J. A. (2003). Chordal graphs and the characteristic polynomial [Electronic Version]. Discrete Mathematics, 262, 211-219. 23. Mielikinen, T. and Ukkonen, E. (2006). The complexity of maxia mum matroid-greedoid intersection and weighted greedoid maximization. Discrete Applied Mathematics, 154(4), 684-691. Retrieved September 5, 2011, from http://arxiv.org/PS cache/CS/pdf/0405/0405094v1. pdf 24. Okamoto, Y. and Nakamura, M. (2003). The forbidden minor characterization of line-search antimatroids of rooted digraphs. Discrete Applied Mathematics, 131(2), 523-533. Retrieved September 4, 2011, from http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.2.6532& rep=rep1&type=pdf 25. Tedford, S. J. (2007). Connectivity of the lifts of a greedoid, The Electronic Journal of Combinatorics, 14, 5pp. http://www.combinatorics. org/Volume 14/PDF/v14i1n9.pdf

23

Anda mungkin juga menyukai