| Popis: |
In this paper we investigate the usefulness of an evolutionary programming (EP) optimization algorithm for tackling the problem of Bandwidth Allocation for Virtual Paths (EP-BAVP). Results for two network topologies obtained with EP are in close agreement with those obtained using a classical constrained optimization (CCO) technique. In the simplest topology (3 nodes) the EP algorithm achieves the same throughputs as the CCO, and converges to the solution within 100 generations, compared to the CCO technique, which converges in 50 iterations. In the more complex topology (4 nodes), the EP algorithm converges within 2000 generations, and the results are very close to the CCO solution which converges in 300 iterations. However, slightly higher throughputs (about 0.5%) were allocated by the classical technique. Furthermore, a similar time span was required for both algorithms for convergence for the 3 and 4 node problems. Our preliminary results suggest that it is worthwhile to obtain further insight into the reported strength of EP-BAVP. |
