Answers are grounded in this paper and cite their sources.
Accurate parameter characterization of dual-polarization optical fibers from limited measurements remains challenging due to nonlinear coupling, sparse observations, and measurement noise. We propose a Numerical Physics-Informed Residual Kolmogorov–Arnold Network (NPR-KAN) for extracting fiber parameters from coherent complex-field measurements acquired at finite propagation positions. NPR-KAN employs a residual Chebyshev-KAN to construct a continuously differentiable representation of the dual-polarization field and incorporates the governing equations as physical constraints. By exploiting the explicit parameter structure of the physical model, sensitivity-weighted regularized variable projection is introduced for numerical parameter inversion, avoiding direct joint optimization of physical parameters and network weights. The method is evaluated on eight analytical propagation states with high-level noise and 50 additional cases generated by the split-step Fourier method. The corresponding average relative errors remain at a low level for the analytical cases and reach 1.03% and 0.38% for the two identified parameters across the 50 numerical cases, demonstrating accurate and stable dual-polarization fiber parameter characterization from sparse coherent measurements.
No takes yet. Share an insight, caveat, or question.
Wang et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: