Norms in Matrix and Vector

Norms in Vector

P-norm

$$||V|{|}_P=(\sum _{i=1}^n|V_i{|}^p{)}^{\frac{1}{p}}$$

Frobenius Norm

$$||V|{|}_F=(\sum _{i=1}^n|V_i{|}^2{)}^{\frac{1}{2}}$$, where in vector it coincides with 2-norm

Infinity and Negative Infinity Norm

$$||V|{|}_{\infty }=\max (|V_i|),i=1,...,n$$

$$||V|{|}_{-\infty }=\min (|V_i|),i=1,...,n$$

Norms in Matrix ($A\in R^{m,n}$)

P-norm

$$\begin{array}{rl}||A|{|}_P& =||\text{vec(A)}|{|}_p\\ & =(\sum _{i=1}^m\sum _{j=1}^n|A_{ij}{|}^p{)}^{\frac{1}{p}}\end{array}$$

Frobenius Norm

$$\begin{array}{rl}||A|{|}_F& =(\sum _{i=1}^m\sum _{j=1}^n|A_{i,j}{|}^2{)}^{\frac{1}{2}}\\ & =\sqrt{\text{trace}(A^{T}A)}\\ & =\sqrt{\sum _{i=1}^{\text{min of m, n}}\text{eigen values of }{A}^{T}A}\\ & =\sqrt{\sum _{i=1}^{\text{min of m,n}}{\sigma }_i^2(A)}\end{array}$$

Nuclear Norm

$$\begin{array}{rl}||A|{|}_{\ast }& =trace(\sqrt{A^{T}A})\\ & =\sum _{i=1}^{\text{min of m, n}}{\sigma }_i(A)\end{array}$$

1-norm

$$||A|{|}_1=\underset{1\le j\le n}{\max }(\sum _{i=1}^m|A_{ij}|)$$

2-norm

$$\begin{array}{rl}||A|{|}_2& ={\sigma }_{max}(A)\\ & =\sqrt{\text{max eigen value of }{A}^{T}A}\\ & \le ||A|{|}_F\\ & =(\sum _{i=1}^m\sum _{j=1}^n|A_{i,j}{|}^2{)}^{\frac{1}{2}}\end{array}$$, $\sigma (A{)}_{\max }$ repesents largest singular value of matrix A.