× Binary Unit Round-off Roundoff Error Number System with base \(\beta\) Floating point error analysis Orders of Convergence Newton's Method Error Analysis for Newton's Method Theorems on Newton's Method Horner's Algorithm Contractive Mapping Polynomial Interpretation Hermite Interpolation Change of Intervals Gaussian Quadrature Error Analysis for Gaussian Quadrature Composite Formulae \(\theta\)-Method Error Initial Value Problem Stability Local Truncation Error for Multistep Methods Schur's Form Hermite Matrix QR Factorization Householder LU Decomposition Norm Positive Definite Gerschgorin Theorem Rayleigh Quotient August 2017 August 2018 January 2019 August 2019 January 2020 August 2020 January 2021 August 2021 January 2022 August 2022 January 2023 August 2023 January 2024 References
☰ Menu

My name is Hanzhang Yin, and I have developed this website as a resource to facilitate the review of key concepts in abstract algebra, and it is not fully completed. Feel free to email me at hanyin@ku.edu if there are any errors or suggestions for improvement.

2019 (August) Numerical Analysis Qualifying Examination

Problem 5. Suppose \( A \in \mathbb{C}^{m \times n} \) with \( m \geq n \).

(a) Show that matrix \( A^*A \) is always Hermitian positive semi-definite.

(b) Show that \( \|A^*A\|_2 = \|A\|_2^2 \).

(c) Show that if \( \text{rank}(A) = n \) then \( A^*A \) is invertible and \( \|A(A^*A)^{-1}A^*\|_2 = 1 \).

(d) Let \( A^\dagger \in \mathbb{C}^{n \times m} \) be the pseudoinverse (the Moore-Penrose inverse) of matrix \( A \) and suppose that \( \text{rank}(A) = n \). Show that \[ A^\dagger = (A^*A)^{-1}A^*. \]

Proof (a). Let \(x\in \mathbb{C}^n\) and we denote \(Ax = y\in \mathbb{C}^m\). Then, we have \[ x^*A^*Ax = (Ax)^*(Ax) = \|Ax\|_2^2 = \|y\|_2^2\geq 0, \] for any \(x\in \mathbb{R}^n\). Thus, we can know that \(A^*A\) is Hermitian positive semi-definite. \( \blacksquare \)

Proof (b). We denote the SVD of \(A\) as \(A = U\Sigma V^*\), where \(U\in \mathbb{C}^{m\times m}\) and \(V\in \mathbb{C}^{n\times n}\) are unitary matrices, and \(\Sigma\in \mathbb{R}^{m\times n}\) is a diagonal matrix with non-negative real numbers on the diagonal. Suppose that \(\Sigma_{1,1}\geq \Sigma_{2,2}\geq \cdots \geq \Sigma_{n,n} \geq 0\). Since the 2-norm of a matrix is the largest singular value of the matrix, we have \(\|A\|_2 = \Sigma_{1,1}\). Besides that, we can know that \(\Sigma_{1, 1}^2 \geq \Sigma_{2, 2}^2 \geq \cdots \geq \Sigma_{n, n}^2 \geq 0\). \[ A^*A = V\Sigma^*U^*U\Sigma V^* = V\Sigma^*\Sigma V^* = V\Sigma\Sigma V^* = V\Sigma^2 V^*. \] Since \(\Sigma_{1, 1}^2\) is the \((1, 1)\)-entry of \(\Sigma^2\), we can know that \(\|\Sigma^2\|_2 = \Sigma_{1, 1}^2\). Then, we can get \[ \|A^*A\|_2 = \|V\Sigma^2 V^*\|_2 = \|\Sigma^2\|_2 = \|\Sigma^2\|_2 = \Sigma_{1, 1}^2. \] Thus, we can know that \[ \|A^*A\|_2 = \|A\|_2^2. \tag*{\(\blacksquare\)} \]