A KNN-Assisted Differential Evolution Algorithm for EM Optimization of Microwave Filters and Antennas

In microwave devices optimization, the full-wave electromagnetic (EM) simulation is usually computational expensive. In this work, to alleviate this problem, a new optimization algorithm is proposed to reduce the total amount of EM simulation cost. The algorithm combines k-nearest-neighbor (KNN) and differential evolution (DE) methods to achieve both convergence speed and the ability of global optimizing. To benchmark the performance of the proposed algorithm, a waveguide filter and an antenna are investigated. The numerical experiments also considered high dimension problems.