NP-SSM

This dataset contains boundary-level training data generated from high-fidelity two-dimensional finite-difference time-domain (FDTD) simulations of electromagnetic wave propagation. Large computational domains equipped with thick perfectly matched layers (PML) were used to approximate unbounded exterior regions and provide reference solutions with minimal artificial reflections. The dataset includes both regular boundary samples and corner samples, with corner cases explicitly oversampled to improve numerical stability. All samples are organized as temporal sequences and stored in compressed NumPy (.npz) format, together with boolean masks indicating edge and corner locations. This dataset is intended for training and evaluating learning-based absorbing or radiation boundary conditions in time-domain wave simulations.