On the optimality and sharpness of Laguerre's lower bound on the smallest eigenvalue of a symmetric positive definite matrix. (Q2402120)
From MaRDI portal
scientific article
Language | Label | Description | Also known as |
---|---|---|---|
English | On the optimality and sharpness of Laguerre's lower bound on the smallest eigenvalue of a symmetric positive definite matrix. |
scientific article |
Statements
On the optimality and sharpness of Laguerre's lower bound on the smallest eigenvalue of a symmetric positive definite matrix. (English)
0 references
6 September 2017
0 references
Let \(A\in\mathbb{C}^{m\times m}\) have real and positive eigenvalues \(\lambda_1\ge\dots\ge\lambda_m\). Applying Laguerre's method to the equation \(\det(\lambda I-A)=0\) with initial value \(\lambda^{(0)}=0\), an increasing sequence \((\lambda^{(n)})\) converging to \(\lambda_m\) is obtained. In particular, \(\lambda^{(1)}\le\lambda_m\), i.e., \[\lambda_m\ge\frac{m}{t_{-1}}\bigg[1+\sqrt{(m-1)\Big(\frac{mt_{-2}}{t_{-1}^2}-1\Big)}\,\bigg]^{-1},\] where \(t_p=\text{tr}\,A^p\). The author presents this bound and proves its optimality (i.e., it is the best possible lower bound for \(\lambda_m\), using only \(m\), \(t_{-1}\) and \(t_{-2}\)).\par \textit{H. Wolkowicz} and \textit{G. P. H. Styan} [Linear Algebra Appl. 29, 471--506 (1980; Zbl 0435.15015)] proved that if \(A\) has real eigenvalues \(\lambda_1\ge\dots\ge\lambda_m\), then \[\lambda_1\le\frac{t_1}{m}+\sqrt{\frac{m-1}{m}\Big(t_2-\frac{t_1^2}{m}\Big)},\] and this bound is optimal. If \(\lambda_m>0\), then these inequalities are equivalent. The author of this paper proves that \[\lambda_m\le\frac{q(q+1)}{t_{-1}}\bigg\{q+\sqrt{q\Big[(q+1)\frac{t_{-2}}{t_{-1}^2}-1\Big]}\,\bigg\}^{-1},\] where \(q\in\mathbb{Z}_+\) satisfies \[q<\frac{t_{-1}^2}{t_{-2}}\le q+1.\]
0 references
eigenvalue bounds
0 references
positive definite matrices
0 references
Laguerre's method
0 references
singular value computation
0 references
dqds algorithm
0 references