Robust Rayleigh quotient minimization and nonlinear eigenvalue problems
نویسندگان
چکیده
In this paper, we study the robust Rayleigh quotient optimization problems that arise when optimizing in worst-case the Rayleigh quotient of data matrices subject to uncertainties. We propose to solve such problem by exploiting its characterization as a nonlinear eigenvalue problem with eigenvector nonlinearity. With this approach, we can show 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 introduced to address this issue: a spectral transformation based on nonlinear shifting and using second-order derivatives. Numerical experiments for applications in data science (generalized eigenvalue classification, common spatial analysis, and linear discriminant analysis) demonstrate the effectiveness of our proposed approaches.
منابع مشابه
A Comparison of Parallel Preconditioners for the Sparse Generalized Eigenvalue Problems by Rayleigh-Quotient Minimization
متن کامل
Nonlinear Rank-one Modification of the Symmetric Eigenvalue Problem
Nonlinear rank-one modification of the symmetric eigenvalue problem arises from eigenvibrations of mechanical structures with elastically attached loads and calculation of the propagation modes in optical fiber. In this paper, we first study the existence and uniqueness of eigenvalues, and then investigate three numerical algorithms, namely Picard iteration, nonlinear Rayleigh quotient iteratio...
متن کاملConvergence orders of iterative methods for nonlinear eigenvalue problems
The convergence analysis of iterative methods for nonlinear eigenvalue problems is in the most cases restricted either to algebraic simple eigenvalues or to polynomial eigenvalue problems. In this paper we consider two classical methods for general holomorphic eigenvalue problems, namely the nonlinear generalized Rayleigh quotient iteration (NGRQI) and the augmented Newton method. For both meth...
متن کاملZentrum für Informationsdienste und Hochleistungsrechnen (ZIH) A primal-dual Jacobi–Davidson-like method for nonlinear eigenvalue problems
We propose a method for the solution of non-Hermitian ill-conditioned nonlinear eigenvalue problems T (λ)x = 0, which is based on singularity theory of nonlinear equations. The singularity of T (λ) is characterized by a scalar condition μ(λ) = 0 where the singularity function μ is implicitly defined by a nonsingular linear system with the appropriately bordered matrix T (λ). One Newton step for...
متن کاملConvergence Analysis of Restarted Krylov Subspace Eigensolvers
The A-gradient minimization of the Rayleigh quotient allows to construct robust and fastconvergent eigensolvers for the generalized eigenvalue problem for (A,M) with symmetric and positive definite matrices. The A-gradient steepest descent iteration is the simplest case of more general restarted Krylov subspace iterations for the special case that all step-wise generated Krylov subspaces are tw...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
عنوان ژورنال:
دوره شماره
صفحات -
تاریخ انتشار 2018