Multidimensional Meta-Model Based Optimization of Optical Transmission Systems

We present a novel meta-model based optimization method for finding the optimum set of system parameters of an optical transmission system. Sampling points from the multidimensional parameter set are selected sparsely by Latin hypercube design, and for these points the signal quality at the receiver is evaluated by numerical simulations. Afterwards an interpolation function of the results is generated, and an optimization algorithm is carried out on this function to locate optima. Regions of interest can be identified and analyzed more precisely by further numerical simulations, which in turn improves the accuracy of the interpolation function. This iterative approach allows to localize the optimum parameter region efficiently with only a small fraction of the computational effort needed for a grid search.