Network reliability analysis and intractability of counting diameter crystal graphs
Resumen:
Consider a stochastic network, where nodes are perfect but links fail independently, ruled by failure probabilities. Additionally, we are given distinguished nodes, called terminals, and a positive integer, called diameter. The event under study is to connect terminals by paths not longer than the given diameter. The probability of this event is called diameter-constrained reliability (DCR, for short). Since the DCR subsumes connectedness probability of random graphs, its computation belongs to the class of NP-Hard problems. The computational complexity for DCR is known for fixed values of the number of terminals k n and diameter d, being n the number of nodes in the network. The contributions of this article are two-fold. First, we extend the computational complexity of the DCR when the terminal size is a function of the number of nodes, this is, when k = k(n). Second, we state counting diameter-critical graphs belongs to the class of NP-Hard problems.
2016 | |
Computational complexity Network reliability Diameter-critical graphs |
|
Inglés | |
Universidad de la República | |
COLIBRI | |
http://hdl.handle.net/20.500.12008/9204 | |
Acceso abierto | |
Licencia Creative Commons Atribución – No Comercial – Sin Derivadas (CC BY-NC-ND 4.0) |
Sumario: | Consider a stochastic network, where nodes are perfect but links fail independently, ruled by failure probabilities. Additionally, we are given distinguished nodes, called terminals, and a positive integer, called diameter. The event under study is to connect terminals by paths not longer than the given diameter. The probability of this event is called diameter-constrained reliability (DCR, for short). Since the DCR subsumes connectedness probability of random graphs, its computation belongs to the class of NP-Hard problems. The computational complexity for DCR is known for fixed values of the number of terminals k n and diameter d, being n the number of nodes in the network. The contributions of this article are two-fold. First, we extend the computational complexity of the DCR when the terminal size is a function of the number of nodes, this is, when k = k(n). Second, we state counting diameter-critical graphs belongs to the class of NP-Hard problems. |
---|