Robust Rayleigh quotient minimization and nonlinear eigenvalue problems
by Z. Bai, D. Lu, B. Vandereycken
Abstract:
We present a theoretical framework for studying the robust Rayleigh quotient optimization problem by exploiting its characterization as a nonlinear eigenvalue problem with eigenvector nonlinearity. Our analysis reveals that a commonly used iterative method can be divergent due to a wrong ordering of the eigenvalues of the corresponding nonlinear eigenvalue problem. Two strategies are proposed to address this issue: a spectral transformation based on nonlinear shifting and using second-order derivatives. Numerical experiments for applications in generalized eigenvalue classification and common spatial analysis demonstrate the effectiveness of our proposed approaches.
Download links:
Reference:
Z. Bai, D. Lu, B. Vandereycken, "Robust Rayleigh quotient minimization and nonlinear eigenvalue problems", In SIAM J. Sci. Comput., vol. 40, no. 5, pp. A3495-A3522, 2018.
Bibtex Entry:
@article{Bai_LV:2018, Abstract = {We present a theoretical framework for studying the robust Rayleigh quotient optimization problem by exploiting its characterization as a nonlinear eigenvalue problem with eigenvector nonlinearity. Our analysis reveals that a commonly used iterative method can be divergent due to a wrong ordering of the eigenvalues of the corresponding nonlinear eigenvalue problem. Two strategies are proposed to address this issue: a spectral transformation based on nonlinear shifting and using second-order derivatives. Numerical experiments for applications in generalized eigenvalue classification and common spatial analysis demonstrate the effectiveness of our proposed approaches.}, Author = {Bai, Z. and Lu, D. and Vandereycken, B.}, Journal = {SIAM J. Sci. Comput.}, Number = {5}, Pages = {A3495--A3522}, Title = {Robust Rayleigh quotient minimization and nonlinear eigenvalue problems}, Volume = {40}, Year = {2018}, Pdf = {http://www.unige.ch/math/vandereycken/papers/published_Bai_LV_2018.pdf}, Doi = {10.1137/18M1167681} }