The d-edge shortest-path problem for a Monge graph
- New Mexico Univ., Albuquerque, NM (United States). Dept. of Computer Science
- California Univ., Riverside, CA (United States). Dept. of Computer Science
- Sandia National Labs.,Albuquerque, NM (United States)
A complete edge-weighted directed graph on vertices 1,2,...,n that assigns cost c(i,j) to the edge (i,j) is called Monge if its edge costs form a Monge array, i.e., for all i < k and j < l, c[i, j]+c[k,l]{le} < c[i,l]+c[k,j]. One reason Monge graphs are interesting is that shortest paths can be computed quite quickly in such graphs. In particular, Wilber showed that the shortest path from vertex 1 to vertex n of a Monge graph can be computed in O(n) time, and Aggarwal, Klawe, Moran, Shor, and Wilber showed that the shortest d-edge 1-to-n path (i.e., the shortest path among all 1-to-n paths with exactly d edges) can be computed in O(dn) time. This paper`s contribution is a new algorithm for the latter problem. Assuming 0 {le} c[i,j] {le} U and c[i,j + 1] + c[i + 1,j] {minus} c[i,j] {minus} c[i + 1, j + 1] {ge} L > 0 for all i and j, our algorithm runs in O(n(1 + 1g(U/L))) time. Thus, when d {much_gt} 1 + 1g(U/L), our algorithm represents a significant improvement over Aggarwal et al.`s O(dn)-time algorithm. We also present several applications of our algorithm; they include length-limited Huffman coding, finding the maximum-perimeter d-gon inscribed in a given convex n-gon, and a digital-signal-compression problem.
- Research Organization:
- Sandia National Labs., Albuquerque, NM (United States)
- Sponsoring Organization:
- USDOE, Washington, DC (United States)
- DOE Contract Number:
- AC04-76DP00789
- OSTI ID:
- 10146169
- Report Number(s):
- SAND-92-1724C; CONF-930194-1; ON: DE93009828
- Resource Relation:
- Conference: 4. annual ACM-SIAM conference on discrete algorithms,Austin, TX (United States),25-27 Jan 1993; Other Information: PBD: 14 Jul 1992
- Country of Publication:
- United States
- Language:
- English
Similar Records
Fully dynamic output bounded single source shortest path problem
Approximation algorithms for curvature-constrained shortest paths