Detecting a hyperbolic quadratic eigenvalue problem by using a subspace algorithm (Q2290930)
From MaRDI portal
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | Detecting a hyperbolic quadratic eigenvalue problem by using a subspace algorithm |
scientific article |
Statements
Detecting a hyperbolic quadratic eigenvalue problem by using a subspace algorithm (English)
0 references
29 January 2020
0 references
The author considers the quadratic eigenvalue problem (QEP): \[ Q(\lambda)x:=(\lambda ^2M+\lambda D+K)x=0 \] where \(M,D,K\) are complex matrices of order \(n\), \(x\neq 0\) is the eigenvector, \(\lambda\) is the corresponding eigenvalue. A Hermitian QEP is hyperbolic if \(M\) is positive definite and \((x^HDx)^2-4(x^HMx)(x^HKx) > 0\) for all nonzero vectors \(x\). There exist many algorithms for detecting hyperbolicity, most of them are not suitable for large QEPs. The paper is devoted to the construction of the following algorithms: Algorithm 1 -- A basic subspace algorithm for detecting a hyperbolic quadratic eigenvalue problem. Algorithm 2 -- Locally optimal block extended conjugate gradient method (LOBeCG) for detecting a hyperbolic QEP. Algorithm 3 -- A basic subspace algorithm for detecting an overdamped QEP. Comparisons of the above algorithms with some other algorithms are considered in Sections 3 and 4. Numerical experiments show that the proposed subspace algorithms can detect very quickly the (non)hyperbolicity of a large Hermitian quadratic eigenvalue problem. Algorithm 2 can be more efficient with preconditioners than without them. The proposed algorithms are supported by detailed codes.
0 references
quadratic eigenvalue problem
0 references
overdamped
0 references
subspace algorithm
0 references
0 references
0 references