Logo
Pattern

Discover published sets by community

Explore tens of thousands of sets crafted by our community.

Matrix Inequalities

15

Flashcards

0/15

Still learning
StarStarStarStar

Cauchy-Schwarz Inequality

StarStarStarStar

For any vectors xx and yy in an inner product space, x,yxy|\langle x, y \rangle| \leq \|x\| \cdot \|y\|. Example: For x=(1,2)x = (1,2) and y=(3,4)y = (3,4), 13+24=1112+2232+42|1 \cdot 3 + 2 \cdot 4| = 11 \leq \sqrt{1^2 + 2^2} \cdot \sqrt{3^2 + 4^2} which simplifies to 11525=1111 \leq \sqrt{5} \cdot \sqrt{25} = 11.

StarStarStarStar

Triangle Inequality for Matrices

StarStarStarStar

For any matrices AA and BB of the same size, A+BA+B\|A + B\| \leq \|A\| + \|B\|. Example: If A=[1234]A = \begin{bmatrix} 1 & 2 \\ 3 & 4 \end{bmatrix} and B=[5678]B = \begin{bmatrix} 5 & 6 \\ 7 & 8 \end{bmatrix}, then A+B\|A+B\| should be less than or equal to A+B\|A\|+\|B\| for any chosen matrix norm.

StarStarStarStar

Sylvester's Criterion

StarStarStarStar

A symmetric matrix AA is positive definite if and only if all leading principal minors of AA are positive. Example: The matrix A=[2112]A = \begin{bmatrix} 2 & 1 \\ 1 & 2 \end{bmatrix} is positive definite because all its principal minors (2, 3) are positive.

StarStarStarStar

Frobenius Norm Inequality

StarStarStarStar

For any matrices AA and BB, ABFAFBF\|A \cdot B\|_F \leq \|A\|_F \cdot \|B\|_F. Example: If AA and BB are 2×22 \times 2 matrices, then the Frobenius norm of their product should not exceed the product of their Frobenius norms.

StarStarStarStar

Schur Product Theorem

StarStarStarStar

For any two positive semidefinite matrices AA and BB, their Schur (or Hadamard) product ABA \circ B is also positive semidefinite. Example: Let AA and BB be 2×22 \times 2 positive semidefinite matrices; their Schur product will have non-negative eigenvalues.

StarStarStarStar

Trace Inequality

StarStarStarStar

If AA is a square matrix and BB is a positive semidefinite matrix, then tr(AB)0\text{tr}(AB) \geq 0. Example: For A=[1234]A = \begin{bmatrix} 1 & 2 \\ 3 & 4 \end{bmatrix} and BB being any positive semidefinite matrix, the trace of their product will be non-negative.

StarStarStarStar

Hadamard's Inequality

StarStarStarStar

For any matrix AA with rows r1,r2,,rnr_1, r_2, \ldots, r_n, the determinant satisfies det(A)r1r2rn|\text{det}(A)| \leq \|r_1\| \cdot \|r_2\| \cdots \|r_n\|. Example: For matrix A=[1234]A = \begin{bmatrix} 1 & 2 \\ 3 & 4 \end{bmatrix}, det(A)=(1)(4)(2)(3)=46=212+2232+42=525=11|\text{det}(A)| = |(1)(4) - (2)(3)| = |4 - 6| = 2 \leq \sqrt{1^2 + 2^2} \cdot \sqrt{3^2 + 4^2} = \sqrt{5} \cdot \sqrt{25} = 11.

StarStarStarStar

von Neumann's Trace Inequality

StarStarStarStar

For any matrices AA and BB of the same size with singular values σ1σ2σn\sigma_1 \geq \sigma_2 \geq \ldots \geq \sigma_n and τ1τ2τn\tau_1 \geq \tau_2 \geq \ldots \geq \tau_n respectively, tr(AB)i=1nσiτi|\text{tr}(A^*B)| \leq \sum_{i=1}^{n}\sigma_i\tau_i. Example: If A=[3001]A = \begin{bmatrix} 3 & 0 \\ 0 & 1 \end{bmatrix} and B=[1003]B = \begin{bmatrix} 1 & 0 \\ 0 & 3 \end{bmatrix}, then tr(AB)=tr(AB)=31+13=6(31+13)=6|\text{tr}(A^*B)| = |\text{tr}(AB)| = |3 \cdot 1 + 1 \cdot 3| = 6 \leq (3 \cdot 1 + 1 \cdot 3) = 6.

StarStarStarStar

Fan Inequality

StarStarStarStar

For Hermitian matrices AA and BB with eigenvalues λ1λ2λn\lambda_1 \geq \lambda_2 \geq \ldots \geq \lambda_n and μ1μ2μn\mu_1 \geq \mu_2 \geq \ldots \geq \mu_n respectively, tr(AB)i=1nλiμi\text{tr}(AB) \leq \sum_{i=1}^{n}\lambda_i\mu_i. Example: If A=[4002]A = \begin{bmatrix} 4 & 0 \\ 0 & 2 \end{bmatrix} and B=[3001]B = \begin{bmatrix} 3 & 0 \\ 0 & 1 \end{bmatrix}, then tr(AB)=43+21=14(43+21)=14\text{tr}(AB) = 4 \cdot 3 + 2 \cdot 1 = 14 \leq (4 \cdot 3 + 2 \cdot 1) = 14.

StarStarStarStar

Positive Semi-Definite Matrix Sum

StarStarStarStar

If AA and BB are positive semidefinite matrices, then A+BA + B is also positive semidefinite. Example: Given that A=[2002]A = \begin{bmatrix} 2 & 0 \\ 0 & 2 \end{bmatrix} and B=[1001]B = \begin{bmatrix} 1 & 0 \\ 0 & 1 \end{bmatrix} are positive semidefinite, their sum [3003]\begin{bmatrix} 3 & 0 \\ 0 & 3 \end{bmatrix} is also positive semidefinite.

StarStarStarStar

Lyapunov's Inequality for Matrices

StarStarStarStar

Given any p,q,r>0p, q, r > 0 such that 1r=1p+1q\frac{1}{r} = \frac{1}{p} + \frac{1}{q}, for matrices AA and BB we have ABrApBq\|AB\|_r \leq \|A\|_p \cdot \|B\|_q. Example: If AA and BB are 2×22 \times 2 matrices and p=q=r=2p = q = r = 2 (Frobenius norm), then AB2A2B2\|AB\|_2 \leq \|A\|_2 \cdot \|B\|_2.

StarStarStarStar

Weyl's Inequality

StarStarStarStar

For Hermitian matrices AA and BB, and their sum A+BA+B, the eigenvalues satisfy λi+j1(A+B)λi(A)+λj(B)\lambda_{i+j-1}(A+B) \leq \lambda_i(A) + \lambda_j(B) for all i,ji, j. Example: If AA and BB are 2×22 \times 2 Hermitian matrices with eigenvalues (λ1,λ2)(\lambda_1, \lambda_2) and (μ1,μ2)(\mu_1, \mu_2) respectively, then λ2(A+B)λ1(A)+μ2(B)\lambda_2(A+B) \leq \lambda_1(A) + \mu_2(B).

StarStarStarStar

Matrix Holder Inequality

StarStarStarStar

For any matrices AA and BB, and p,q>1p, q > 1 with 1p+1q=1\frac{1}{p} + \frac{1}{q} = 1, we have tr(AB)ApBq|\text{tr}(AB^*)| \leq \|A\|_p \cdot \|B\|_q. Example: For p=q=2p = q = 2 (Frobenius norms), tr(AB)AFBF|\text{tr}(AB^*)| \leq \|A\|_F \cdot \|B\|_F for any A,BA, B.

StarStarStarStar

Oppenheim's Inequality

StarStarStarStar

For positive semidefinite matrices AA and BB such that BB is invertible, det(A+B)det(A)+det(B)\text{det}(A+B) \geq \text{det}(A) + \text{det}(B). Example: Let A=[2002]A = \begin{bmatrix} 2 & 0 \\ 0 & 2 \end{bmatrix} and B=[1004]B = \begin{bmatrix} 1 & 0 \\ 0 & 4 \end{bmatrix} then det(A+B)=det[3006]=18det(A)+det(B)=4+4=8\text{det}(A+B) = \text{det}\begin{bmatrix} 3 & 0 \\ 0 & 6 \end{bmatrix} = 18 \geq \text{det}(A) + \text{det}(B) = 4 + 4 = 8.

StarStarStarStar

Minkowski's Inequality for Matrices

StarStarStarStar

For matrices AA and BB under any unitarily invariant norm, A+BA+B\|A+B\| \leq \|A\| + \|B\|. Example: If AA and BB are 2×22 \times 2 matrices and using the Frobenius norm, then A+BFAF+BF\|A+B\|_F \leq \|A\|_F + \|B\|_F, which holds as an equality if and only if AA and BB are positive multiples of each other.

Know
0
Still learning
Click to flip
Know
0
Logo

© Hypatia.Tech. 2024 All rights reserved.