Operator-probability adaptation, in a genetic-algorithm/heuristic hybrid for minimum cost routing and wavelength allocation of multi-wavelength all-optical transport networks is described. The hybrid algorithm uses an object-oriented representation of networks, and incorporates four operators: path mutation, single-point crossover, reroute and shift-out. The adaptation algorithm is based on that by Davis, but uses simplified operator accounting. Experimental results from three fifteen-node test networks, obtained using a tool for optical network optimisation, modelling and design (NOMaD), illustrate the interesting dynamic behaviour of the adaptation algorithm. They suggest that, in this application, with powerful problem-specific operators, the main benefits of operator-probability adaptation are in relieving the experimenter of the burden of setting initial probabilities and in the early performance of the hybrid, rather than in improvements of the final solution quality obtained.
|Title of host publication||1998 IEEE International Conference on Evolutionary Computation Proceedings. IEEE World Congress on Computational Intelligence (Cat. No.98TH8360)|
|Publication status||Published - May 1998|