In a packet radio network (PRN), transmission of packets involves both the transmission of information the end user needs and the transmission of additional or "overhead" information necessary for proper operation of the network. Two solutions to the multicast transmission problem have been presented, one which employs the Hopfield network, and the other based on heuristic algorithms. Although both approaches usually yield suboptimal performance when compared to an exhaustive search, the time required by these methods to reach a solution is lower by several orders of magnitude for all but the simplest PRNs. The Hopfield method is extremely easy to implement and converged to a valid solution in over 98% of our simulations. This is considered a high success rate for this type of network, which we attribute to the sparseness of the configuration and careful choice of parameters. The heuristic is an improvement over the neural network yielding the lowest number of transmissions and requiring the least amount of time. Either approach may be suitable for determining solutions to this NP-complete problem depending on the actual application and computational resources available.
[1]
S. S. Sengupta,et al.
The traveling salesman problem
,
1961
.
[2]
Faouzi Kamoun,et al.
Neural networks for shortest path computation and routing in computer networks
,
1993,
IEEE Trans. Neural Networks.
[3]
S. E. Dreyfus,et al.
The steiner problem in graphs
,
1971,
Networks.
[4]
Mischa Schwartz,et al.
Telecommunication networks: protocols, modeling and analysis
,
1986
.
[5]
Mahesan Niranjan,et al.
A theoretical investigation into the performance of the Hopfield model
,
1990,
IEEE Trans. Neural Networks.