Repository logo
 

Faster graph algorithms via switching classes

Date

2012

Authors

Lindzey, Nathan, author
McConnell, Ross, advisor
Bohm, Wim, committee member
Penttila, Tim, committee member

Journal Title

Journal ISSN

Volume Title

Abstract

The runtime of an algorithm is intimately related to how an instance is represented. Recall that the runtimes of the first generation of graph algorithms were expressed as functions of n := |V|. This analysis was natural since at this time graphs were represented in n2 space via their adjacency matrix. It was soon noticed that if m := |E| = o(n2), then a variety of graph algorithms could be sped-up by computing the adjacency-list from the adjacency matrix, then running the algorithm on the more efficient adjacency-list representation. This motivated the introduction of m to the runtime of graph algorithms and it is now customary in algorithm design to assume that a graph instance is given in the form of its adjacency-list. For instance, a graph algorithm is not considered to run in linear time unless it runs in O(n + m) time. An O(n2) bound is not considered linear, even though the two bounds are the same in the worst case. Let m͂ be the size of the minimum representative of a graph G's switching class (w.r.t. to some switching operation). It is shown that better bounds for several classical graph algorithms can be obtained by modifying them so that their running time is a function of n+m͂ rather than of n+m. This is significant because m͂ is O(m) but m is not O(m͂). This is accomplished by first computing the so-called partially complemented adjacency list (pc-list) from an adjacency list, then designing an algorithm that is amenable to the more efficient pc-list representation. The pc-list data-structure is generalization of the adjacency list that has a natural correspondence to switching classes. Using this approach, better bounds are obtained for bipartite maximum matching, graph diameter, and vertex-weighted all-pairs shortest path.

Description

Rights Access

Subject

algorithms
dense graphs
graph compression
graph representation
graph theory
switching

Citation

Associated Publications