Regular ArticleAlgorithms for Degree Constrained Graph Factors of Minimum Deficiency
Abstract
If H is a graph and g and ƒ are integer valued functions on the vertex set of H, then a (g, ƒ)-factor of H is a subgraph G of H whose degree at each vertex v of H lies in the interval [g(v), ƒ(v)]. Thus, a (0,1)-factor of H is just a matching of H, and a (1, 1)-factor of H is a perfect matching of H. In case H admits no (g, ƒ)-factor, one may wish to find a subgraph which is "as close as possible" to being a (g, ƒ)-factor. We consider only subgraphs which are themselves (0, ƒ)-factors of H (these are clearly guaranteed to exist), and distinguish two ways in which to measure the extent to which a (0, ƒ)-factor falls short of being a (g, ƒ)-factor, i.e., to which it fails to meet the lower bounds g(v) on the degrees of v. The first way involves the notion of g-deficiency, which is defined to be the sum of all amounts by which the degrees of the vertices fall short of g(v). Our main result is an efficient algorithm for the construction of a (0, ƒ)-factor of H of minimum g-deficiency. When H is bipartite (or more generally, when g is uniformly less than ƒ except on a set of vertices that induces a bipartite subgraph of H), we obtain an ([formula](||E|| + g(V))) algorithm by a non-trivial modification of the Hopcroft-Karp bipartite matching algorithm. (Here ||E|| denotes the number of non-parallel edges of H and g(V) is the sum of all lower bounds g(v), v ∈ V.) Denoting by p the number of vertices v with g(v) = ƒ(v), we derive an ([formula](||E|| + g(V)) + p||E||) algorithm for general H. What distinguishes our algorithms from earlier work (such as the ([formula]|E|) algorithm of Gabow) is that they highlight the role played by the function g in the complexity of the problem; when g is substantially smaller than ƒ our algorithms provide substantially tighter upper bounds. We establish a linear equivalence between the maximum flow problem for 0-1 networks and the minimum g-deficiency problem in the bipartite graphs. This equivalence allows us to derive the best known complexity bounds on the various cases of the maximum flow problem from our algorithm. Somewhat surprizingly, it is not clear how to derive our complexity bounds from the best maximum flow algorithms. The second way to measure how well a (0, ƒ)-factor approximates a (g, ƒ)-factor by the number of vertices v with degree at least g(v). This quantity, called the g-saturation, is to be maximised. We show that this problem is NP-hard, provided g ≥ 3.
References (0)
Cited by (24)
Subgraphs with orthogonal factorizations and algorithms
2010, European Journal of CombinatoricsLet be a graph and let and be two integer-valued functions defined on such that for every . Let be vertex-disjoint subgraphs of with (). In this paper, we prove that every -graph contains a subgraph such that has a -factorization orthogonal to (), where and are positive integers with . We also prove that there are polynomial-time algorithms for finding the desired subgraph .
Linear-time certifying algorithms for near-graphical sequences
2009, Discrete MathematicsA sequence of non-negative integers is graphical if it is the degree sequence of some graph, that is, there exists a graph on vertices whose th vertex has degree , for . The notion of a graphical sequence has a natural reformulation and generalization in terms of factors of complete graphs.
If is a graph and and are integer-valued functions on the vertex set , then a -factor of is a subgraph of whose degree at each vertex lies in the interval . Thus, a -factor is just a matching of and a (1, 1)-factor is a perfect matching of . If is complete then a -factor realizes a degree sequence that is consistent with the sequence of intervals .
Graphical sequences have been extensively studied and admit several elegant characterizations. We are interested in extending these characterizations to non-graphical sequences by introducing a natural measure of “near-graphical”. We do this in the context of minimally deficient -factors of complete graphs. Our main result is a simple linear-time greedy algorithm for constructing minimally deficient -factors in complete graphs that generalizes the method of Hakimi and Havel (for constructing -factors in complete graphs, when possible). It has the added advantage of producing a certificate of minimum deficiency (through a generalization of the Erdös–Gallai characterization of -factors in complete graphs) at no additional cost.
Greedily constructing maximal partial f-factors
2009, Discrete MathematicsLet be a graph and let be a function . A partial -factor of is a subgraph of , such that the degree in of every vertex is at most . We study here the recognition problem of graphs, where all maximal partial -factors have the same number of edges. Graphs which satisfy that property for the function are known as equimatchable and their recognition problem is the subject of several previous articles in the literature. We show the problem is polynomially solvable if the function is bounded by a constant, and provide a structural characterization for graphs with girth at least 5 in which all maximal partial 2-factors are of the same size.
Some problems on factorizations with constraints in bipartite graphs
2003, Discrete Applied MathematicsLet G=(X,Y,E(G)) be a bipartite graph with vertex set V(G)=X∪Y and edge set E(G) and let g and f be two non-negative integer-valued functions defined on V(G) such that g(x)⩽f(x) for each x∈V(G). A (g,f)-factor of G is a spanning subgraph F of G such that g(x)⩽dF(x)⩽f(x) for each x∈V(F); a (g,f)-factorization of G is a partition of E(G) into edge-disjoint (g,f)-factors. In this paper it is proved that every bipartite (mg+m−1,mf−m+1)-graph has (g,f)-factorizations randomly k-orthogonal to any given subgraph with km edges if k⩽g(x) for any x∈V(G) and has a (g,f)-factorization k-orthogonal to any given subgraph with km edges if k−1⩽g(x) for any x∈V(G) and that every bipartite (mg,mf)-graph has a (g,f)-factorization orthogonal to any given m-star if 0⩽g(x)⩽f(x) for any x∈V(G). Furthermore, it is shown that there are polynomial algorithms for finding the desired factorizations and the results in this paper are in some sense best possible.
Extremal subgraphs with respect to vertex degree bounds
2003, Discrete MathematicsA weighted graph (G,w) is a graph G=(V,E) together with a weight function on its vertices . We consider the maximal (minimal) number of edges in a spanning subgraph H of G with for all v∈V. For these parameters and their corresponding duals we prove—among other results—generalizations of the well-known theorems of König (Mat. Fiz. Lapok 38 (1931) 116–119) and Gallai (Ann. Univ. Sci. Budapest. Eötvös Sect. Math. 2 (1959) 133–138). Note that these notions naturally generalize maximum matchings and minimum edge covers in graphs.
Minus domination in small-degree graphs
2001, Discrete Applied MathematicsMinus domination in graphs is a variant of domination where the vertices must be labeled −1,0,+1 such that the sum of labels in each N[v] is positive. (As usual, N[v] means the set containing v together with its neighbors.) The minus domination number γ− is the minimum total sum of labels that can be achieved. In this paper we prove linear lower bounds for γ− in graphs either with Δ⩽3, or with Δ⩽4 but without vertices of degree 2. The central section is concerned with complexity results for Δ⩽4: We show that computing γ− is NP-hard and MAX SNP-hard there, but that γ− can be approximated in linear time within some constant factor. Finally, our approach also applies to signed domination (where the labels are −1,+1 only) in small-degree graphs.