부산대학교 도서관

Expanding on the research on physics-informed neural networks (PINNs) to solve the eigenvalue problems in black hole (BH) perturbation theory, the supervised learning approach was investigated to solve the Regge-Wheeler and Teukolsky equations governing gravitational perturbations of Schwarzschild and Kerr BHs. Previous works have applied unsupervised PINNs to compute quasinormal mode frequencies, i.e. the discrete spectra of eigenvalues that satisfy the BH perturbation equations; however, that was limited to only the zeroth and first overtone (i.e. n = 0, 1). In this paper, supervised learning is used to compute higher overtones with approximation errors accentuated primarily by the input data rather than the PINNs themselves. Our results show the ability of PINNs to approximate overtones up to the limits set by the input data, which concurs with the universal approximation theory of neural networks.
Comment: 22 pages, 3 figures