Electromagnetics optimization using an evolutionary algorithm with a mixed-parameter self-adaptive mutation operator

Applications of evolutionary programming (EP) in electromagnetics to date have been mainly in continuous parameter optimizations. EP, however, can directly work with continuous or discrete parameters. We present an implementation of EP with a mixed continuous-discrete parameter representation. In our approach the mutation operator consists of a hybrid combination of Gaussian mutation, for the continuous parameters, and Poisson mutation, for the discrete parameters. The implementation uses self-adaptive schemes for updating the standard deviation of the Gaussian distribution and the mean of the Poisson distribution during the evolution. As an example, the mixed-parameter EP algorithm is applied to the design of a multi-layer filter structure.