Linear Algebra

We collect here some Linear Algebra concepts that appears here and there in these notes.

In all the following discussion the symbol \(\mathbb{K}\) can be read as either the set of real numbers \(\mathbb{R}\), or \(\mathbb{C}\) the set of complex numbers.

Definition 8

An \(m \times p\) matrix over \(\mathbb{K}\) is an \(m \times p\) rectangular array \(A = (a_{i,j})_{\substack{i=1,\ldots,m\\j=1,\ldots,p}}\) with entries in \(\mathbb{K}\). The size of \(A\) is \(m\times p\) and \(A\) is said to be square if \(m = p\).

Two matrices \(A\) and \(B\) are said to be equal if they have the same size and \(a_{i,j} = b_{i,j}\) \(\forall \,i,j\). We can build \(\mathbb{K}^{m\times p}\) as the set of all the matrices of size \(m\times p\) over the field \(\mathbb{K}\) and define an addition and a scalar multiplication over \(\mathbb{K}\) by \(A+B \triangleq (a_{i,j}+b_{i,j})_{i,j}\) and the product with a scalar \(c\in\mathbb{K}\) as \(c A = (c a_{i,j})_{i,j}\), for all \(c \in \mathbb{K}\).

Definition 9

Let \(A = (a_{i,j}) \in \mathbb{K}^{m\times p}\) and let \(\mathbf{b} \in \mathbb{K}^{p\times 1} \equiv \mathbb{K}^{p}\). The matrix–vector product of \(A\) and \(\mathbf{b}\) is the vector \(\mathbf{c}=A\mathbf{b}\) given elementwise by \(c_i= \sum_{j=1}^{p} \mathbf{a}_{i,j} b_j\).

Let \(A \in \mathbb{K}^{m\times p}\), \(B \in \mathbb{K}^{p\times n}\) we define the matrix product of \(A\) and \(B=[\mathbf{b}_{:,1},\ldots,\mathbf{b}_{:,n}]\) as \(AB = [A\mathbf{b}_{:,1},\ldots,A\mathbf{b}_{:,n}]\) and, elementwise as

\[ C=AB, \quad C=(c_{i,j}), \quad c_{i,j}=\sum_{k=1}^p a_{i,k} b_{k,j}.\]

The other operations that are interesting to look at for matrices and vectors are norms. In an abstract setting, norms are particular type of functions obeying the following characterization.

Definition 10

Let \((V,\mathbb{K})\) be a vector space with \(\mathbb{K} = \mathbb{R}\) or \(\mathbb{C}\) and \(f:V\rightarrow \mathbb{R}\) a function such that

  1. \(f(\mathbf{v}) \geq 0\) \(\forall \, \mathbf{v} \in V\) and such that \(f(\mathbf{v}) = 0\) if and only if \(\mathbf{v}=\mathbf{0}\);

  2. \(f(\alpha \mathbf{v}) = |\alpha|f(\mathbf{v})\), \(\forall\,\alpha \in\mathbb{K}\), \(\forall\, \mathbf{v} \in V\);

  3. \(f(\mathbf{v}+\mathbf{w}) \leq f(\mathbf{v})+f(\mathbf{w})\) \(\forall \, \mathbf{v},\mathbf{w} \in V\).

The function \(f\) is a norm for the vector space \(V\). The triple \((V,\mathbb{K},f)\) is called a normed space.

Then examples of such norms for vectors are given by

Theorem 1

Let \(V = \mathbb{C}^n\) or \(\mathbb{R}^n\). If \(p \geq 1\) we have that

\[\|\mathbf{x}\|_p \triangleq \left(\sum_{i=1}^{n}|x_i|^p\right)^{\frac{1}{p}}, \;\; \forall\, \mathbf{x} \in V,\]

is a norm. Moreover,

\[\|\mathbf{x}\|_{\infty} \triangleq \max_i |x_i| = \lim_{p \rightarrow \infty} \|\mathbf{x}\|_p,\]

is a norm as well.

In general one can define also norm for matrices, e.g.,

Definition 11

Given a matrix \(A \in \mathbb{K}^{m \times n}\), with \(\mathbb{K} = \mathbb{R}\) or \(\mathbb{C}\), the Frobenius norm of \(A\) is defined as,

\[\|A\|_F \triangleq \left(\sum_{i=1}^{m} \sum_{j=1}^{n} |a_{i,j}|^2\right)^{\frac{1}{2}}.\]

The matrix norms associated to the \(p\)–norms from Theorem 1 can be defined also as

\[\|A\|_p \triangleq \max_{\mathbf{v} \neq \mathbf{0}} \frac{\|A \mathbf{v} \|_p }{\| \mathbf{v} \|_p}.\]

Note that

\[\|A\|_1 \triangleq \max_{1 \leq j \leq n} \sum_{i=1}^m |a_{i,j}|, \quad \|A\|_\infty \triangleq \max_{1 \leq 1 \leq m} \sum_{j=1}^n |a_{i,j}|.\]

previous

Working with images