Discussion Overview
The discussion revolves around estimating complex eigenvalues of a real matrix A. Participants explore various numerical methods and algorithms suitable for this task, particularly in the context of the QR algorithm and alternatives for computing eigenvalues.
Discussion Character
- Technical explanation
- Debate/contested
- Mathematical reasoning
Main Points Raised
- One participant notes that while the QR algorithm works for real eigenvalues, it fails to converge for complex eigenvalues and seeks alternatives.
- Another participant suggests using the LAPACK routine dgeev.f, which does not require complex arithmetic and can handle complex eigenvalues as conjugate pairs, although they mention the need to reduce the matrix to Hessenberg form first.
- A third participant discusses the QR algorithm's limitations in generating eigenvalue estimates for complex conjugate pairs and proposes two methods for handling this: calculating eigenvalues from a 2x2 matrix or using a "Wilkinson shift" to navigate the complex plane.
- One participant introduces Gershgorin's Circle Theorem as a potential method if the off-diagonal entries of the matrix are small, although they express uncertainty about its relevance.
- A separate query arises regarding the estimation of eigenvalues for a matrix with complex elements, indicating a need for subroutines that do not require double precision.
Areas of Agreement / Disagreement
Participants present multiple competing views on the best approach to estimate complex eigenvalues, and the discussion remains unresolved with no consensus on a single method.
Contextual Notes
Some methods discussed depend on specific conditions, such as the size of off-diagonal entries or the requirement for complex arithmetic. The effectiveness of the proposed methods may vary based on the properties of the matrix in question.