Radial basis function-assisted global + local optimization algorithm for polarization-maintaining hollow-core anti-resonant fiber.

Abstract

Traditional evolutionary algorithms require a large number of computations for fiber optimization. This paper proposes an enhanced radial basis function-based global and local optimization algorithm to improve efficiency. The algorithm incorporates a surrogate model-based pre-screening module, which integrates both global and local models to improve the exploration and exploitation of the search space. It also employs a local model-assisted search strategy to refine promising regions. A comparative analysis with several other optimization methods is provided. To optimize two low-loss, polarization-maintaining fiber structures, the algorithm needs just 100 evaluations per model, decreasing the computational effort by at least an order of magnitude compared to conventional methods. The best-achieved birefringence is 1 × 10^-4, with a loss of 0.3 dB/m.