Surrogate model-assisted gradient descent optimization for low-loss, high-birefringence hollow-core anti-resonant fiber

This paper proposes an improved structure for low-loss polarization-maintaining fibers and employs a combination of gradient descent algorithm and radial basis function surrogate model for parameter optimization. The loss and birefringence are optimized using the weighted sum method, with comparisons made for different weights. Numerical experiments demonstrate the method’s excellent optimization efficiency.