Study Guides / Linear Algebra / Exercise session 8
EXERCISE SESSION 8 · ETH ZURICH · HS19 · linalg

Linear Maps and Change of Basis

Linear map, endomorphism, automorphism, image, kernel, commutative diagram, equivalent/similar matrices

SB
Steven Battilana Quant · ex-ETH

Recap (Orthogonal and Unitary Matrices)

Definition

A complex n×nn \times n matrix AA is called unitary if AHA=AAH=𝟙A^HA = AA^H = \text{𝟙}.
A real n×nn \times n matrix AA is called orthogonal if ATA=AAT=𝟙A^TA = AA^T = \text{𝟙}.

Satz

If A,B∈En×nA,B \in \mathbb{E}^{n \times n} are unitary (or orthogonal) matrices, then:

  1. AA is regular
  2. A−1=AHA^{-1} = A^H (or A−1=ATA^{-1} = A^T)
  3. A−1A^{-1} is unitary (orthogonal)
  4. ABAB is unitary (orthogonal)
Definition

The Kronecker delta is defined by:

δij={1,i=j0,i≠j\delta_{ij} = \begin{cases} 1, \quad i=j\\ 0, \quad i \neq j \end{cases}
Definition unit vectors
e1:=(10⋮0),e2:=(010⋮0),ei:=(0⋮010⋮0),en:=(0⋮01)e_1 := \begin{pmatrix} 1\\ 0\\ \vdots\\ 0 \end{pmatrix}, e_2 := \begin{pmatrix} 0\\ 1\\ 0\\ \vdots\\ 0 \end{pmatrix}, e_i := \begin{pmatrix} 0\\ \vdots\\ 0\\ 1\\ 0\\ \vdots\\ 0 \end{pmatrix}, e_n := \begin{pmatrix} 0\\ \vdots\\ 0\\ 1 \end{pmatrix}
Example 1
  • ⟨e1,e2⟩=0\langle e_1, e_2 \rangle = 0
  • ⟨e1,e1⟩=1\langle e_1, e_1 \rangle = 1
Definition orthonormal

Let a,b∈Ena,b \in \mathbb{E}^n. The vectors a,ba,b are orthonormal if the following conditions are satisfied:

  1. The vectors are normalised, i.e.: ∥a∥=1and∥b∥=1.\Vert a \Vert = 1 \quad \text{and} \quad \Vert b \Vert = 1.
  2. The vectors are orthogonal, i.e.: ⟨a,b⟩={1,a=b0,a≠b\langle a, b \rangle = \begin{cases} 1, & a=b\\ 0, & a \neq b \end{cases}
Remark

For an orthogonal matrix A∈En×nA \in \mathbb{E}^{n \times n} of the form A=(a1∣...∣an)A = (a_1 | ... | a_n), the column vectors a1,...,ana_1, ..., a_n are pairwise orthonormal.

Linear Maps

Definition

A map f:X→Yf: X \to Y is called injective if

∀x1,x2∈X:x1≠x2⇒f(x1)≠f(x2).\forall x_1,x_2 \in X: \quad x_1 \neq x_2 \quad \Rightarrow \quad f(x_1) \neq f(x_2).

(In words: different elements of XX are mapped to different images in YY.)

Definition

A map f:X→Yf: X \to Y is called surjective if

∀y∈Y ∃x∈X:f(x)=y.\forall y \in Y \, \exists x \in X: \quad f(x) = y.

(In words: every element of YY is "hit" by ff.)

Definition

A map f:X→Yf: X \to Y is called bijective if

∀y∈Y ∃!x∈X:f(x)=y.\forall y \in Y \, \exists \boldsymbol{!} x \in X: \quad f(x) = y.

(In words: every element of YY is "hit" by ff exactly once.)

Definition

A map F:V→WF: V \to W between E\mathbb{E}-vector spaces VV and WW is called linear (more precisely a homomorphism of E\mathbb{E}-vector spaces) if ∀v,w∈V\forall v,w \in V and ∀λ∈E\forall \lambda \in \mathbb{E}:

L1 F(v+w)=F(v)+F(w)\quad F(v+w) =F(v)+F(w)
L2 F(λv)=λF(v)\quad F(\lambda v) = \lambda F(v)

These two conditions can be combined into one:

L F(v+λw)=F(v)+λF(w)\quad F(v +\lambda w) = F(v) + \lambda F(w).

Remark notation

For F:V→WF: V \to W linear, F∈Hom(V,W)F \in \text{Hom}(V,W).

Remark

It is common to sharpen the notion of homomorphism:

  1. F∈Hom(V,W)F \in \text{Hom}(V,W) and bijective ⇔\quad \Leftrightarrow \quad isomorphism (notation: V=~WV \widetilde{=} W)
  2. F∈Hom(V,W)F \in \text{Hom}(V,W) and V=W⇔V=W \quad \Leftrightarrow \quad endomorphism (notation: F∈End(V)F \in End(V))
  3. F∈End(V)F \in \text{End}(V) and bijective ⇔\quad \Leftrightarrow \quad automorphism

Moreover: \quad (i) ⇔∃G:W→V\Leftrightarrow \exists G: W \to V linear such that F∘G=idWF \circ G = \text{id}_W, G∘F=idVG \circ F = \text{id}_V, i.e.

∀w∈W:F(G(w))=w∀v∈V:G(F(v))=v\begin{aligned} \forall w \in W: \quad &F(G(w)) = w\\ \forall v \in V: \quad &G(F(v)) = v \end{aligned}
Remark

Let M(F), M(G)M(F),\, M(G) be the representation matrices of F:V→WF: V \to W isomorphic and its inverse map G=F−1:W→VG = F^{-1}: W \to V and let V,WV,W be finite-dimensional vector spaces, i.e. dim(V)<∞\text{dim}(V) \lt \infty and dim(W)<∞\text{dim}(W) \lt \infty. Then bijectivity of FF means that

  • dim(V)=dim(W)\text{dim}(V) = \text{dim}(W)
  • M(F)⋅M(G)=M(G)⋅M(F)=𝟙dim(V)⇔M(F)=(M(G))−1M(F) \cdot M(G) = M(G) \cdot M(F) = \text{𝟙}_{\text{dim}(V)} \Leftrightarrow M(F) = (M(G))^{-1}
Definition

For every basis B={v1,...,vn}\mathcal{B} = \{v_1, ..., v_n\} of VV there is exactly one isomorphism:

ϕB:En→V,(x1,...,xn)↦ϕB(x1,...,xn)=∑k=1nxkvk=x1v1+...+xnvn with ϕB(ei)=vi.\phi_B: \mathbb{E}^n \to V, (x_1, ..., x_n) \mapsto \phi_B(x_1, ..., x_n) = \sum_{k=1}^n x_k v_k = x_1 v_1 + ... + x_n v_n \text{ with } \phi_B(e_i) = v_i.

(In words: ϕB\phi_B assigns to xx its coordinates with respect to the basis B\mathcal{B}.)

Definition

Let F∈Hom(V,W)F \in \text{Hom}(V,W).

  • Im(F):=F(V)={F(v)∣v∈V}⊂W\text{Im}(F) := F(V) = \{F(v) | v \in V\} \subset W is a subspace of WW and is called image(F)\text{image}(F) or Im(F)\mathbf{Im}\boldsymbol{(F)}.
  • ker(F):={v∈V∣F(v)=0}⊂V\text{ker}(F) := \{v \in V | F(v) = 0\} \subset V is a subspace of VV and is called ker(F)\mathbf{ker}\boldsymbol{(F)}.
Satz

Let F:V→WF: V \to W be linear and V, W vector spaces. Then:

  1. F(0)=0F(0) = 0, zero is always mapped to zero
  2. F surjective ⇔Im(F)=W⇔dim(Im(F))=dim(W)\Leftrightarrow \text{Im}(F) = W \Leftrightarrow \text{dim}(\text{Im}(F)) = \text{dim}(W)
  3. F injective ⇔ker(F)={0}⇔dim(ker(F))=0\Leftrightarrow \text{ker}(F) = \{\boldsymbol{0}\} \Leftrightarrow \text{dim}(\text{ker}(F)) = 0
  4. FF is an isomorphism ⇔dim(V)=dim(W)=rank(F)\quad \Leftrightarrow \quad \text{dim}(V) = \text{dim}(W) = \text{rank}(F)
Satz

If f,gf,g are linear maps ⇒f∘g\quad \Rightarrow f \circ g is a linear map.

Satz

If f,gf,g are linear maps, then the function F:=f±gF:= f \pm g arising from the linear combination of f,gf,g is again a linear map.

Definition

The rank of the linear map FF is defined as:

rank(F)=dim(Im(F)).\text{rank}(F) = \text{dim}(\text{Im}(F)).
Remark

The rank of the linear map FF equals the rank of its representation matrix M(F)M(F). We have: rank(F)=dim(Im(F))=rank(M(F))=rank(M(F)T)\quad\quad \text{rank}(F) = \text{dim}(\text{Im}(F)) = \text{rank}(M(F)) = \text{rank}(M(F)^T)

Remark

Row rank = column rank: rank(M(F))=rank(M(F)T)\quad \text{rank}(M(F)) = \text{rank}(M(F)^T)
Caution: in general: \quad column space ≠\neq row space

Satz

Let VV and WW be two finite-dimensional subspaces of a larger vector space (finite-dimensional ⇔dim(V)=n<∞\Leftrightarrow \text{dim}(V)=n\lt\infty and dim(W)=k<∞\text{dim}(W)=k\lt\infty) and let f:V→Wf: V \to W be linear; then the following dimension formulas hold:

  • dim(V+W)=dim(V)+dim(W)−dim(V∩W)\text{dim}(V+W) = \text{dim}(V) + \text{dim}(W)-\text{dim}(V \cap W)
  • n=dim(V)=dim(ker(f))+dim(Im(f))n = \text{dim}(V) = \text{dim}(\text{ker}(f)) + \text{dim}(\text{Im}(f))
Remark properties of linear maps

Let V,WV, W be E\mathbb{E}-vector spaces and B={v1,...,vn}\mathcal{B} = \{v_1, ..., v_n\} a basis of VV. Let F:V→WF: V \to W be linear.

  • Im(F)=span(F(v1),...,F(vn))\text{Im}(F) = \text{span}(F(v_1), ..., F(v_n)), i.e. FF is uniquely defined by its values on the basis vectors
  • If FF is injective and v1,...,vn∈Vv_1, ..., v_n \in V are linearly independent, then F(v1),...,F(vn)∈Im(F)F(v_1), ..., F(v_n) \in \text{Im}(F) are linearly independent
  • dim(V)=dim(W)<∞\text{dim}(V) = \text{dim}(W) \lt \infty and FF injective ⇒F\quad \Rightarrow \quad F is bijective!
Example 2

Let F:R→R,x↦5x−1F: \mathbb{R} \to \mathbb{R}, x \mapsto 5x-1.

FF is not linear, since F(0)=−1≠0F(0) = -1 \neq 0.

Remark

Let f∈Hom(V,W)f \in \text{Hom}(V,W) and let the vector spaces V,W be finite-dimensional. The image of ff is spanned by the columns of M(f)M(f), i.e.
Im(f)=span{columns of M(f)}\text{Im}(f) = \text{span}\{\text{columns of } M(f)\}.

Remark

Let F∈Hom(V,W)F \in \text{Hom}(V,W) with corresponding representation matrix M(F)M(F). Then:

F(x)=0⇔M(F)x=0.F(x) = 0 \quad \Leftrightarrow \quad M(F)x = 0.
Example 3

Given:

F:R3→R3,x=(x1x2x3)↦(x1+x2−x33x1+x2+2x32x1+3x3)F: \mathbb{R}^3 \to \mathbb{R}^3, \quad x = \begin{pmatrix} x_1\\ x_2\\ x_3 \end{pmatrix} \mapsto \begin{pmatrix} x_1 + x_2 - x_3\\ 3x_1 + x_2 + 2x_3\\ 2x_1 + 3x_3 \end{pmatrix}with M(F)=(11−1312203),s.t.M(F)x=F(x)\text{with } M(F) = \begin{pmatrix} 1 & 1 & -1\\ 3 & 1 & 2\\ 2 & 0 & 3 \end{pmatrix}, \text{s.t.} \quad M(F)x = F(x)

Wanted: ker(F)\text{ker}(F), a basis of ker(F)\text{ker}(F), Im(F)\text{Im}(F), a basis of Im(F)\text{Im}(F)

To compute ker(F)\text{ker}(F), compute the row echelon form of M(F)M(F):

M(F)=(11−1312203)⇝(ii)−l21(i)(11−10−25203)⇝(iii)−l31(i)(11−10−250−25)M(F) = \begin{pmatrix} 1 & 1 & -1\\ 3 & 1 & 2\\ 2 & 0 & 3 \end{pmatrix} \stackrel{(ii)-l_{21}(i)}{\leadsto} \begin{pmatrix} 1 & 1 & -1\\ 0 & -2 & 5\\ 2 & 0 & 3 \end{pmatrix} \stackrel{(iii)-l_{31}(i)}{\leadsto} \begin{pmatrix} 1 & 1 & -1\\ 0 & -2 & 5\\ 0 & -2 & 5 \end{pmatrix}⇝(iii)−l32(ii)(11−10−25000)⇝(i)−1−2(ii)(10320−25000)⇝2(i)(2030−25000)=:M~(F)\stackrel{(iii)-l_{32}(ii)}{\leadsto} \begin{pmatrix} 1 & 1 & -1\\ 0 & -2 & 5\\ 0 & 0 & 0 \end{pmatrix} \stackrel{(i)-\frac{1}{-2}(ii)}{\leadsto} \begin{pmatrix} 1 & 0 & \frac{3}{2}\\ 0 & -2 & 5\\ 0 & 0 & 0 \end{pmatrix} \stackrel{2(i)}{\leadsto} \begin{pmatrix} 2 & 0 & 3\\ 0 & -2 & 5\\ 0 & 0 & 0 \end{pmatrix} =: \widetilde{M}(F)

To find a basis we use the rank-nullity theorem (dimension formula):

dim(R3)=dim(ker(F))+dim(Im(F))⇔dim(ker(F))=dim(R3)−dim(Im(F))⇔dim(ker(F))=3−2=1\begin{aligned} \text{dim}(\mathbb{R}^3) &= \text{dim}(\text{ker}(F)) + \text{dim}(\text{Im}(F))\\ \Leftrightarrow \quad\quad \text{dim}(\text{ker}(F)) &= \text{dim}(\mathbb{R}^3) - \text{dim}(\text{Im}(F))\\ \Leftrightarrow \quad\quad \text{dim}(\text{ker}(F)) &= 3 - 2\\ &= 1 \end{aligned}

We can read off dim(Im(F))=rank(F)=rank(M(F))=2\text{dim}(\text{Im}(F)) = \text{rank}(F) = \text{rank}(M(F)) = 2 directly from M~(F)\widetilde{M}(F). Thanks to the dimension formula we know that the basis of the kernel of FF contains one basis vector.

Below we compute the solution set of M(F)M(F) (or M~(F)\widetilde{M}(F)), which is at the same time ker(F)=ker(M(F))=ker(M~(F))\text{ker}(F)=\text{ker}(M(F))=\text{ker}(\widetilde{M}(F)), since we are solving the linear system M(F)x=0M(F)x=0.
(As you saw in the exercise session, we can also "guess" the vectors in the kernel with the help of M~(F)\widetilde{M}(F).)

M~(F)⇒row 3t:=x3,    t∈Rfree parameterM~(F)⇒row 2−2x2+5x3=0⇔t=x3−2x2+5t=0⇔2x2=5t⇔x2=52t\begin{aligned} \widetilde{M}(F) \quad &\stackrel{\text{row 3}}{\Rightarrow} \quad t:=x_3, \;\; t \in \mathbb{R} \quad \text{free parameter}\\ \widetilde{M}(F) \quad &\stackrel{\text{row 2}}{\Rightarrow} \quad -2x_2+5x_3 = 0\\ &\stackrel{t=x_3}{\Leftrightarrow} \quad -2x_2+5t = 0\\ &\Leftrightarrow \quad 2x_2 = 5t\\ &\Leftrightarrow \quad x_2 = \frac{5}{2}t \end{aligned}M~(F)⇒row 12x1+3x3=0⇔t=x32x1+3t=0⇔2x1=−3t⇔x1=−32t\begin{aligned} \widetilde{M}(F) \quad &\stackrel{\text{row 1}}{\Rightarrow} \quad 2x_1+3x_3 = 0\\ &\stackrel{t=x_3}{\Leftrightarrow} \quad 2x_1+3t = 0\\ &\Leftrightarrow \quad 2x_1 = -3t\\ &\Leftrightarrow \quad x_1 = -\frac{3}{2}t \end{aligned}⟹L={(x1x2x3)∣(−32521)⋅t,t∈R}\Longrightarrow \quad\quad \mathbb{L} = \left\{ \begin{pmatrix} x_1\\ x_2\\ x_3 \end{pmatrix} \left| \begin{pmatrix} -\frac{3}{2}\\ \frac{5}{2}\\ 1 \end{pmatrix} \cdot t, t \in \mathbb{R} \right. \right\}⇒ker(F)=ker(M(F))={(x1x2x3)∣(−32521)⋅t,  t∈R}=span{(x1x2x3)∣(−32521)}=span{(x1x2x3)∣(−352)}\begin{aligned} \Rightarrow \text{ker}(F) = \text{ker}(M(F)) &= \left\{ \begin{pmatrix} x_1\\ x_2\\ x_3 \end{pmatrix} \left| \begin{pmatrix} -\frac{3}{2}\\ \frac{5}{2}\\ 1 \end{pmatrix} \cdot t,\,\, t \in \mathbb{R} \right. \right\}\\ &= \text{span}\left\{ \begin{pmatrix} x_1\\ x_2\\ x_3 \end{pmatrix} \left| \begin{pmatrix} -\frac{3}{2}\\ \frac{5}{2}\\ 1 \end{pmatrix} \right. \right\}\\ &= \text{span}\left\{ \begin{pmatrix} x_1\\ x_2\\ x_3 \end{pmatrix} \left| \begin{pmatrix} -3\\ 5\\ 2 \end{pmatrix} \right. \right\} \end{aligned}⇒A basis of ker(F) is for example:Bker(F)={(−352)}\Rightarrow \text{A basis of ker}(F) \text{ is for example}: \mathcal{B}_{\text{ker}(F)} = \left\{ \begin{pmatrix} -3\\ 5\\ 2 \end{pmatrix} \right\}

For the image Im(F)\text{Im}(F) we know from the remark above:

Im(F)=span{(132),(110),(−123)}\text{Im}(F) = \text{span}\left\{ \begin{pmatrix} 1\\ 3\\ 2 \end{pmatrix}, \begin{pmatrix} 1\\ 1\\ 0 \end{pmatrix}, \begin{pmatrix} -1\\ 2\\ 3 \end{pmatrix} \right\}

From the columns of M(F)M(F) (a spanning set of Im(F)\text{Im}(F)), choose two vectors as a basis of Im(F)\text{Im}(F), namely those that have pivot elements in M~(F)\widetilde{M}(F):

⇒BIm(F)={(132),(110)}\Rightarrow \mathcal{B}_{\text{Im}(F)} = \left\{ \begin{pmatrix} 1\\ 3\\ 2 \end{pmatrix}, \begin{pmatrix} 1\\ 1\\ 0 \end{pmatrix} \right\}
Example 4

Given: F:R3→R2F: \mathbb{R}^3 \to \mathbb{R}^2 is given by the following matrix:

M(F)=(123456)M(F) = \begin{pmatrix} 1 & 2 & 3\\ 4 & 5 & 6 \end{pmatrix}

Wanted: determine bases of ker(F)\text{ker}(F) and Im(F)\text{Im}(F).

M(F)=(123456)⇝(ii)−l21(i)(1230−3−6)⇝(i)−2−3(ii)(10−10−3−6)=:M~(F)M(F) = \begin{pmatrix} 1 & 2 & 3\\ 4 & 5 & 6 \end{pmatrix} \stackrel{(ii) - l_{21}(i)}{\leadsto} \begin{pmatrix} 1 & 2 & 3\\ 0 & -3 & -6 \end{pmatrix} \stackrel{(i) - \frac{2}{-3} (ii)}{\leadsto} \begin{pmatrix} 1 & 0 & -1\\ 0 & -3 & -6 \end{pmatrix} =: \widetilde{M}(F)

To find dim(ker(F))\text{dim}(\text{ker}(F)) we use the rank-nullity theorem (dimension formula):

dim(R3)=dim(ker(F))+dim(Im(F))⇔dim(ker(F))=dim(R3)−dim(Im(F))⇔dim(ker(F))=3−2=1\begin{aligned} \text{dim}(\mathbb{R}^3) &= \text{dim}(\text{ker}(F)) + \text{dim}(\text{Im}(F))\\ \Leftrightarrow \quad\quad \text{dim}(\text{ker}(F)) &= \text{dim}(\mathbb{R}^3) - \text{dim}(\text{Im}(F))\\ \Leftrightarrow \quad\quad \text{dim}(\text{ker}(F)) &= 3 - 2\\ &= 1 \end{aligned}

We can read off dim(Im(F))=rank(F)=rank(M(F))=2\text{dim}(\text{Im}(F)) = \text{rank}(F) = \text{rank}(M(F)) = 2 directly from M~(F)\widetilde{M}(F). Thanks to the dimension formula we know that the basis of the kernel of FF contains one basis vector.

⇒ker(F)=ker(M(F))={(x1x2x3)∣(1−21)⋅t,  t∈R}=span{(x1x2x3)∣(1−21)}\begin{aligned} \Rightarrow \text{ker}(F) = \text{ker}(M(F)) &= \left\{ \begin{pmatrix} x_1\\ x_2\\ x_3 \end{pmatrix} \left| \begin{pmatrix} 1\\ -2\\ 1 \end{pmatrix} \cdot t,\,\, t \in \mathbb{R} \right. \right\}\\ &= \text{span}\left\{ \begin{pmatrix} x_1\\ x_2\\ x_3 \end{pmatrix} \left| \begin{pmatrix} 1\\ -2\\ 1 \end{pmatrix} \right. \right\} \end{aligned}⇒A basis of ker(F) is for example:Bker(F)={(1−21)}\Rightarrow \text{A basis of ker}(F) \text{ is for example}: \mathcal{B}_{\text{ker}(F)} = \left\{ \begin{pmatrix} 1\\ -2\\ 1 \end{pmatrix} \right\}

For the image Im(F)\text{Im}(F) we know from the remark above:

Im(F)=span{(14),(25),(36)}\text{Im}(F) = \text{span}\left\{ \begin{pmatrix} 1\\ 4 \end{pmatrix}, \begin{pmatrix} 2\\ 5 \end{pmatrix}, \begin{pmatrix} 3\\ 6 \end{pmatrix} \right\}

From the columns of M(F)M(F) (a spanning set of Im(F)\text{Im}(F)), choose two vectors as a basis of Im(F)\text{Im}(F), namely those that have pivot elements in M~(F)\widetilde{M}(F):

⇒BIm(F)={(14),(25)}\Rightarrow \mathcal{B}_{\text{Im}(F)} = \left\{ \begin{pmatrix} 1\\ 4 \end{pmatrix}, \begin{pmatrix} 2\\ 5 \end{pmatrix} \right\}
Algorithm matrix of a linear map (representation matrix; special case with standard basis)
  1. Given: V,WV, W vector spaces, F:(V,A)→(W,B),v↦F(v)F: (V, \mathcal{A}) \to (W, \mathcal{B}), v \mapsto F(v) and a basis of VV with A={a1,...,an}\mathcal{A} = \{a_1, ..., a_n\} and the standard basis of WW with B={b1,...,bm}\mathcal{B} = \{b_1, ..., b_m\}
  2. Wanted: MBA(F)M_{\mathcal{B}}^{\mathcal{A}}(F)
  3. Compute F(ai),  i∈{1,...,n}F(a_i), \,\, i \in \{1, ..., n\} for every basis vector
  4. Build MBA(F)=(F(a1),...,F(an)⏟n columns)(}m rowsM_{\mathcal{B}}^{\mathcal{A}}(F) = (\underbrace{F(a_1), ..., F(a_n)}_{n\text{ columns}}) \left. \vphantom{\Big(} \right\} \quad m \text{ rows}.
  5. We have obtained the matrix of FF, where the domain is given with respect to A\mathcal{A} and the codomain with respect to B\mathcal{B}.
Example 5

Let F=ddt:P2→P1,p↦p˙=dpdtF= \frac{d}{dt}: \mathcal{P}_2 \to \mathcal{P}_1, p \mapsto \dot{p} = \frac{dp}{dt}.

(i) To show: FF is a linear map.

Proof: ∀a,b∈P2,  λ∈E:\forall a,b \in \mathcal{P}_2, \,\, \lambda \in \mathbb{E}:

F(a+λb)=ddt(a+λb)=ddta+λddtb=F(a)+λF(b)□\begin{aligned} F(a+\lambda b) &= \frac{d}{dt} (a+\lambda b)\\ &= \frac{d}{dt} a + \lambda \frac{d}{dt} b\\ &= F(a) + \lambda F(b) \quad\quad\quad\quad \square \end{aligned}

(ii) Find the matrix MBB(F)M_{\mathcal{B}}^{\mathcal{B}}(F) with respect to the monomial basis B={1,t,t2}\mathcal{B} = \{1, t, t^2\}.

p=λ0+λ1t+λ2t2∈P2,p=^(λ0λ1λ2)∈E3p = \lambda_0 + \lambda_1 t + \lambda_2 t^2 \in \mathcal{P}_2, \quad p \mathrel{\hat{=}} \begin{pmatrix} \lambda_0\\ \lambda_1\\ \lambda_2 \end{pmatrix} \in \mathbb{E}^3
p˙=λ1+2λ2t∈P1,p˙=^(λ12λ20)∈E3\dot{p} = \lambda_1 + 2\lambda_2 t \in \mathcal{P}_1, \quad \dot{p} \mathrel{\hat{=}} \begin{pmatrix} \lambda_1\\ 2\lambda_2\\ 0 \end{pmatrix} \in \mathbb{E}^3

1=^e1t=^e2t2=^e3\begin{aligned} 1 &\mathrel{\hat{=}} e_1\\ t &\mathrel{\hat{=}} e_2\\ t^2 &\mathrel{\hat{=}} e_3 \end{aligned}p˙(1)=^p˙(e1)=(000),p˙(t)=^p˙(e2)=(100),p˙(t2)=^p˙(e3)=(020)\dot{p}(1) \mathrel{\hat{=}} \dot{p}(e_1) = \begin{pmatrix} 0\\ 0\\ 0 \end{pmatrix}, \quad \dot{p}(t) \mathrel{\hat{=}} \dot{p}(e_2) = \begin{pmatrix} 1\\ 0\\ 0 \end{pmatrix}, \quad \dot{p}(t^2) \mathrel{\hat{=}} \dot{p}(e_3) = \begin{pmatrix} 0\\ 2\\ 0 \end{pmatrix}⟹MBB(F)=(010002000)\Longrightarrow \quad\quad\quad M_{\mathcal{B}}^{\mathcal{B}}(F)= \begin{pmatrix} 0 & 1 & 0\\ 0 & 0 & 2\\ 0 & 0 & 0 \end{pmatrix}
Remark

A polynomial p∈Pnp \in \mathcal{P}_n is uniquely determined by its function values p(xi)p(x_i) at n+1n+1 pairwise distinct points xix_i, i∈{0,...,n}i \in \{0, ..., n\}.

Remark

Let V,U,W  EV, U, W\,\, \mathbb{E}-vector spaces with dim(V)=n\text{dim}(V) = n, dim(U)=k\text{dim}(U) = k and dim(W)=ℓ\text{dim}(W) = \ell; then the dimension rule for compositions of linear maps is:

M(f)∈Rℓ×n, M(g)∈Rk×ℓ⇒M(g∘f)=M(g)⋅M(f)∈("Rk×ℓ⋅Rℓ×n")=Rk×nM(f) \in \mathbb{R}^{\ell \times n},\, M(g) \in \mathbb{R}^{k \times \ell} \quad \Rightarrow \quad M(g \circ f) = M(g) \cdot M(f) \in (\text{"}\mathbb{R}^{k \times \ell} \cdot \mathbb{R}^{\ell \times n}\text{"}) = \mathbb{R}^{k \times n}

Change of Basis, Coordinate Transformation

Definition

Let A=(e1,...,en)\mathcal{A} = (e_1, ..., e_n) be the canonical basis of the vector space V and B=(b1,...,bn)\mathcal{B} = (b_1, ..., b_n) another basis of V, described in terms of the canonical basis. Then there exists a change-of-basis matrix with:

TAB=(b1∣⋯∣bn) with bi=TABei, i∈{1,...,n}.T_A^B = (b_1 | \cdots | b_n) \quad \text{ with } b_i = T_A^B e_i, \, i \in \{1, ..., n\}.VA→TBAVBV_\mathcal{A} \xrightarrow{\quad T_B^A \quad} V_\mathcal{B}
Remark

The following calculation rule holds: TBA=(TAB)−1\quad T_B^A = (T_A^B)^{-1}
With the definition above we therefore obtain: TBAbi=ei, i∈{1,...,n}\quad T_B^A b_i = e_i, \, i \in \{1, ..., n\}.

Satz

Let E\mathbb{E} be a field, VV an E\mathbb{E}-vector space with dim(V)=n<∞\text{dim}(V) = n \lt \infty. Let v∈Vv \in V and A={a1,...,an},  B={b1,...,bn}\mathcal{A} = \{a_1, ..., a_n\},\,\, \mathcal{B} = \{b_1, ..., b_n\} bases of VV. Then there exist unique λ1,...,λn∈E\lambda_1, ..., \lambda_n \in \mathbb{E} and unique μ1,...,μn∈E\mu_1, ..., \mu_n \in \mathbb{E} such that

v=∑k=1nλkak=∑k=1nμkbk.v = \sum_{k=1}^n \lambda_k a_k = \sum_{k=1}^n \mu_k b_k.

This raises the question of how to switch between the bases A\mathcal{A} and B\mathcal{B}; concretely, you are for example given the matrix of a linear map with respect to A\mathcal{A} and now want to express the matrix with respect to B\mathcal{B}.

Definition

For every basis B={v1,...,vn}\mathcal{B} = \{v_1, ..., v_n\} of VV there is exactly one isomorphism:

ϕB:En→V,(x1,...,xn)↦ϕB(x1,...,xn)=∑k=1nxkvk=x1v1+...+xnvn with ϕB(ei)=vi.\phi_B: \mathbb{E}^n \to V, (x_1, ..., x_n) \mapsto \phi_B(x_1, ..., x_n) = \sum_{k=1}^n x_k v_k = x_1 v_1 + ... + x_n v_n \text{ with } \phi_B(e_i) = v_i.

(In words: ϕB\phi_B assigns to xx its coordinates with respect to the basis B\mathcal{B}.)

Definition

Let VV with basis A={v1,...,vm}\mathcal{A}=\{v_1, ..., v_m\} and WW with basis B={w1,...,wn}\mathcal{B}=\{w_1, ..., w_n\} be vector spaces over E\mathbb{E}. Then for every linear map f:V→Wf: V \to W there is exactly one matrix MBA(f)M_B^A(f) such that MBA(f)j=f(vj)=a1jw1+...+anjwnM_B^A(f)_j = f(v_j) = a_{1j}w_1 + ... + a_{nj}w_n for j=1,...,mj = 1, ..., m.

Remark

The jj-th column of the matrix MBA(f)M_B^A(f) above is the coordinate vector of f(vj)f(v_j) with respect to the basis B\mathcal{B}.

Remark important

The columns of a matrix of a linear map contain the images of the basis vectors, i.e. MBA(f)=(f(v1)∣⋯∣f(vm))M_B^A(f) = (f(v_1)| \cdots |f(v_m)).

Remark

Using the commutative diagram, the matrix MBA(f)M_B^A(f) can also be described as follows:

MBA(f)=ϕB−1∘f∘ϕAM_B^A(f) = \phi_B^{-1} \circ f \circ \phi_AEm→MBA(f)EnϕA↓↓ϕBV→fW\begin{CD} \mathbb{E}^m @>{M_B^A(f)}>> \mathbb{E}^n \\ @V{\phi_A}VV @VV{\phi_B}V \\ V @>{f}>> W \end{CD}
Definition

The regular change-of-basis matrix TBAT_{\mathcal{B}}^{\mathcal{A}} with bases A={v1,...,vn}, B={w1,...,wn}\mathcal{A}=\{v_1, ..., v_n\}, \, \mathcal{B} = \{w_1, ..., w_n\} of the vector space V looks as follows:

TBA=ϕB−1ϕA=(t11⋯t1n⋮⋮tn1⋯tnn),T_{\mathcal{B}}^{\mathcal{A}} = \phi_B^{-1} \phi_A = \begin{pmatrix} t_{11} & \cdots & t_{1n}\\ \vdots && \vdots\\ t_{n1} & \cdots & t_{nn} \end{pmatrix},
En\mathbb{E}^nEn\mathbb{E}^nVV TBAT_B^AϕA\phi_AϕB\phi_B

This lets you now describe wi=ti1v1+...+tinvn=(TBAvA)i,i∈{1,...,n}w_i = t_{i1}v_1 + ... + t_{in}v_n = (T_{\mathcal{B}}^{\mathcal{A}} \boldsymbol{v}_{\mathcal{A}})_i, \quad i \in \{1, ...,n\}, where wiw_i is expressed with respect to B\mathcal{B} and viv_i with respect to A\mathcal{A}:

TBAvA=wB,where vA=(v1⋮vn) w.r.t. A,  wB=(w1⋮wn) w.r.t. B.T_{\mathcal{B}}^{\mathcal{A}} \boldsymbol{v}_{\mathcal{A}} = \boldsymbol{w}_{\mathcal{B}}, \quad \text{where } \boldsymbol{v}_{\mathcal{A}} = \begin{pmatrix} v_1\\ \vdots\\ v_n \end{pmatrix} \text{ w.r.t. } \mathcal{A},\,\, \boldsymbol{w}_{\mathcal{B}} = \begin{pmatrix} w_1\\ \vdots\\ w_n \end{pmatrix} \text{ w.r.t. } \mathcal{B}.
Remark calculation rules
  • TAA=𝟙T_{\mathcal{A}}^{\mathcal{A}} = \text{𝟙}

  • TBA=(TAB)−1T_{\mathcal{B}}^{\mathcal{A}} = (T_{\mathcal{A}}^{\mathcal{B}})^{-1}

  • λA∈Kn\lambda_{\mathcal{A}} \in \mathbb{K}^n a coordinate vector with respect to A\mathcal{A}
    μB∈Kn\mu_{\mathcal{B}} \in \mathbb{K}^n a coordinate vector with respect to B\mathcal{B}
    ⇒TBAλA=μB\Rightarrow \quad T_{\mathcal{B}}^{\mathcal{A}} \lambda_{\mathcal{A}} = \mu_{\mathcal{B}}

  • f:V→Vf: V \to V linear with matrix MAA(f)M_A^A(f), where the domain and the codomain are given with respect to AA. Analogously, the matrix MBB(f)M_B^B(f) is given with respect to BB in both the domain and the codomain. We achieve a change of basis from AA to BB of the matrix MAA(f)M_A^A(f) with the change-of-basis matrices TAB,  TBAT_A^B,\,\, T_B^A:

    MBB(f)=TBAMAA(f)TABM_B^B(f) = T_B^A M_A^A(f) T_A^B
    En\mathbb{E}^nEn\mathbb{E}^nVVVVEn\mathbb{E}^nEn\mathbb{E}^n MBB(f)M_B^B(f)ffMAA(f)M_A^A(f)ϕB\phi_BϕB\phi_BϕA\phi_AϕA\phi_ATABT_A^BTBAT_B^A
  • f:V→Vf: V \to V linear with matrix MB2B1(f)M_{B_2}^{B_1}(f), where the domain is given with respect to B1B_1 and the codomain with respect to B2B_2. Analogously, the matrix MB2′B1′(f)M_{B_2'}^{B_1'}(f) is given with respect to B1′B_1' in the domain and B2′B_2' in the codomain. We achieve a change of basis from B1B_1 to B1′B_1' (domain) and from B2B_2 to B2′B_2' (codomain) of the matrix MB2B1(f)M_{B_2}^{B_1}(f) with the change-of-basis matrices TB2′B2,  TB1B1′T_{B_2'}^{B_2},\,\, T_{B_1}^{B_1'}:

    MB2′B1′(f)=TB2′B2MB2B1(f)TB1B1′M_{B_2'}^{B_1'}(f) = T_{B_2'}^{B_2} M_{B_2}^{B_1}(f) T_{B_1}^{B_1'}
    En\mathbb{E}^nEn\mathbb{E}^nVVVVEn\mathbb{E}^nEn\mathbb{E}^n MB2′B1′(f)M_{B_2'}^{B_1'}(f)ffMB2B1(f)M_{B_2}^{B_1}(f)ϕB1′\phi_{B_1'}ϕB2′\phi_{B_2'}ϕB1\phi_{B_1}ϕB2\phi_{B_2}TB1B1′T_{B_1}^{B_1'}TB2′B2T_{B_2'}^{B_2}
Algorithm change-of-basis matrix
  1. Given: A=(a1,...,an), B=(b1,...,bn)\mathcal{A} = (a_1, ..., a_n),\, \mathcal{B} = (b_1, ..., b_n) are bases of V.
  2. Wanted: change-of-basis matrices TBAT_B^A and TABT_A^B.
  3. (BA)⇔(b1⋯bna1⋯an)⇝⋯⇝Gaussian elimination without row swaps(𝟙TBA)\displaystyle \left(\begin{array}{c|c} B & A \end{array}\right) \Leftrightarrow \left(\begin{array}{ccc|ccc} b_1 & \cdots & b_n & a_1 & \cdots & a_n \end{array}\right) \stackrel{\text{Gaussian elimination without row swaps}}{\leadsto \cdots \leadsto} \left(\begin{array}{c|c} \text{𝟙} & T_B^A \end{array}\right)
  4. (AB)⇔(a1⋯anb1⋯bn)⇝⋯⇝Gaussian elimination without row swaps(𝟙TAB)\displaystyle \left(\begin{array}{c|c} A & B \end{array}\right) \Leftrightarrow \left(\begin{array}{ccc|ccc} a_1 & \cdots & a_n & b_1 & \cdots & b_n \end{array}\right) \stackrel{\text{Gaussian elimination without row swaps}}{\leadsto \cdots \leadsto} \left(\begin{array}{c|c} \text{𝟙} & T_A^B \end{array}\right)
Remark intuition
(BA)⇝(B−1B⏟𝟙1B−1A⏟=TBA)⇝(𝟙TBA)\left(\begin{array}{c|c} B & A \end{array}\right) \leadsto \left(\begin{array}{c|c} \underbrace{B^{-1}B}_{\text{𝟙}\vphantom{1}} & \underbrace{B^{-1}A}_{= T_B^A} \end{array}\right) \leadsto \left(\begin{array}{c|c} \text{𝟙} & T_B^A \end{array}\right)
Example 6

Given: A={(110),(101),(011)},B={(334),(123),(345)}\mathcal{A} = \left\{ \begin{pmatrix} 1\\ 1\\ 0 \end{pmatrix}, \begin{pmatrix} 1\\ 0\\ 1 \end{pmatrix}, \begin{pmatrix} 0\\ 1\\ 1 \end{pmatrix} \right\}, \mathcal{B} = \left\{ \begin{pmatrix} 3\\ 3\\ 4 \end{pmatrix}, \begin{pmatrix} 1\\ 2\\ 3 \end{pmatrix}, \begin{pmatrix} 3\\ 4\\ 5 \end{pmatrix} \right\}
Wanted: TBA,TABT_B^A, T_A^B

(313110324101435011)\left(\begin{array}{ccc|ccc} 3 & 1 & 3 & 1 & 1 & 0\\ 3 & 2 & 4 & 1 & 0 & 1\\ 4 & 3 & 5 & 0 & 1 & 1 \end{array}\right)⇝(ii)−l21(i)  (3131100110−11435011)⇝(iii)−l31(i)  (3131100110−110531−43−131)⇝(iii)−l32(ii)  (3131100110−1100−23−4343−23)⇝(iii)⋅−32  (3131100110−110012−21)⇝(ii)−(iii)  (313110010−2100012−21)⇝(i)−3⋅(iii)  (310−57−3010−2100012−21)⇝(i)−(ii)  (300−36−3010−2100012−21)⇝(i)⋅13  (100−12−1010−2100012−21)\begin{aligned} \stackrel{(ii) - l_{21}(i)}{\leadsto} \; & \left(\begin{array}{ccc|ccc} 3 & 1 & 3 & 1 & 1 & 0\\ 0 & 1 & 1 & 0 & -1 & 1\\ 4 & 3 & 5 & 0 & 1 & 1 \end{array}\right)\\[1.5ex] \stackrel{(iii) - l_{31}(i)}{\leadsto} \; & \left(\begin{array}{ccc|ccc} 3 & 1 & 3 & 1 & 1 & 0\\ 0 & 1 & 1 & 0 & -1 & 1\\ 0 & \frac{5}{3} & 1 & -\frac{4}{3} & -\frac{1}{3} & 1 \end{array}\right)\\[1.5ex] \stackrel{(iii) - l_{32}(ii)}{\leadsto} \; & \left(\begin{array}{ccc|ccc} 3 & 1 & 3 & 1 & 1 & 0\\ 0 & 1 & 1 & 0 & -1 & 1\\ 0 & 0 & -\frac{2}{3} & -\frac{4}{3} & \frac{4}{3} & -\frac{2}{3} \end{array}\right)\\[1.5ex] \stackrel{(iii) \cdot -\frac{3}{2}}{\leadsto} \; & \left(\begin{array}{ccc|ccc} 3 & 1 & 3 & 1 & 1 & 0\\ 0 & 1 & 1 & 0 & -1 & 1\\ 0 & 0 & 1 & 2 & -2 & 1 \end{array}\right)\\[1.5ex] \stackrel{(ii) - (iii)}{\leadsto} \; & \left(\begin{array}{ccc|ccc} 3 & 1 & 3 & 1 & 1 & 0\\ 0 & 1 & 0 & -2 & 1 & 0\\ 0 & 0 & 1 & 2 & -2 & 1 \end{array}\right)\\[1.5ex] \stackrel{(i) - 3 \cdot (iii)}{\leadsto} \; & \left(\begin{array}{ccc|ccc} 3 & 1 & 0 & -5 & 7 & -3\\ 0 & 1 & 0 & -2 & 1 & 0\\ 0 & 0 & 1 & 2 & -2 & 1 \end{array}\right)\\[1.5ex] \stackrel{(i) - (ii)}{\leadsto} \; & \left(\begin{array}{ccc|ccc} 3 & 0 & 0 & -3 & 6 & -3\\ 0 & 1 & 0 & -2 & 1 & 0\\ 0 & 0 & 1 & 2 & -2 & 1 \end{array}\right)\\[1.5ex] \stackrel{(i) \cdot \frac{1}{3}}{\leadsto} \; & \left(\begin{array}{ccc|ccc} 1 & 0 & 0 & -1 & 2 & -1\\ 0 & 1 & 0 & -2 & 1 & 0\\ 0 & 0 & 1 & 2 & -2 & 1 \end{array}\right) \end{aligned}⟹TBA=(−12−1−2102−21)\Longrightarrow \quad\quad\quad\quad T_B^A = \begin{pmatrix} -1 & 2 & -1\\ -2 & 1 & 0\\ 2 & -2 & 1 \end{pmatrix}
Remark

You can compute TAB=(TBA)−1T_A^B = (T_B^A)^{-1} either with the recipe above, or you use the recipe from exercise session 4 and compute the inverse (TBA)−1=TAB(T_B^A)^{-1} = T_A^B.

Example 7

Let V=PV = \mathcal{P} with bases B={1,x,x2}\mathcal{B} = \{1, x, x^2\} standard basis (monomial basis) and A={a1,a2,a3}\mathcal{A} = \{a_1, a_2, a_3\} with

a1=x2a2=(x+1)2=x2+2x+1a3=(x−1)2=x2−2x+1\begin{aligned} a_1 &= x^2\\ a_2 &= (x+1)^2 = x^2 + 2x +1\\ a_3 &= (x-1)^2 = x^2-2x +1 \end{aligned}

a) TBA,TABT_B^A, T_A^B?

b) Let F=ddt:P2→P1,p↦p˙=dpdtF= \frac{d}{dt}: \mathcal{P}_2 \to \mathcal{P}_1, p \mapsto \dot{p} = \frac{dp}{dt} with matrix
M(F)=MBB(F)=(010002000)M(F) = M_B^B(F) = \begin{pmatrix} 0 & 1 & 0\\ 0 & 0 & 2\\ 0 & 0 & 0 \end{pmatrix}.
What is MAA(F)M_A^A(F)?

c) Let p(x)=3x2−8x+2∈P2p(x) = 3x^2 - 8x +2 \in \mathcal{P}_2. What are the coordinates of pp with respect to A\mathcal{A} and B\mathcal{B}?

a) Since B\mathcal{B} is the standard basis, we have: TBA=(a1∣a2∣a3)=(01102−2111)T_B^A = (a_1 | a_2 | a_3) = \begin{pmatrix} 0 & 1 & 1\\ 0 & 2 & -2\\ 1 & 1 & 1 \end{pmatrix}
TAB=(TBA)−1:T_A^B = (T_B^A)^{-1}:

(01110002−2010111001)⇝row swaps(11100101110002−2010)\left(\begin{array}{ccc|ccc} 0 & 1 & 1 & 1 & 0 & 0\\ 0 & 2 & -2 & 0 & 1 & 0\\ 1 & 1 & 1 & 0 & 0 & 1 \end{array}\right) \stackrel{\text{row swaps}}{\leadsto} \left(\begin{array}{ccc|ccc} 1 & 1 & 1 & 0 & 0 & 1\\ 0 & 1 & 1 & 1 & 0 & 0\\ 0 & 2 & -2 & 0 & 1 & 0 \end{array}\right)⇝(iii)−l32(ii)(11100101110000−4−210)⇝(ii)−1−4(iii)(1110010101214000−4−210)\stackrel{(iii) - l_{32}(ii)}{\leadsto} \left(\begin{array}{ccc|ccc} 1 & 1 & 1 & 0 & 0 & 1\\ 0 & 1 & 1 & 1 & 0 & 0\\ 0 & 0 & -4 & -2 & 1 & 0 \end{array}\right) \stackrel{(ii) - \frac{1}{-4}(iii)}{\leadsto} \left(\begin{array}{ccc|ccc} 1 & 1 & 1 & 0 & 0 & 1\\ 0 & 1 & 0 & \frac{1}{2} & \frac{1}{4} & 0\\ 0 & 0 & -4 & -2 & 1 & 0 \end{array}\right)⇝(i)−1−4(iii)(110−121410101214000−4−210)⇝(i)−(ii)(100−1010101214000−4−210)\stackrel{(i) - \frac{1}{-4}(iii)}{\leadsto} \left(\begin{array}{ccc|ccc} 1 & 1 & 0 & -\frac{1}{2} & \frac{1}{4} & 1\\ 0 & 1 & 0 & \frac{1}{2} & \frac{1}{4} & 0\\ 0 & 0 & -4 & -2 & 1 & 0 \end{array}\right) \stackrel{(i) - (ii)}{\leadsto} \left(\begin{array}{ccc|ccc} 1 & 0 & 0 & -1 & 0 & 1\\ 0 & 1 & 0 & \frac{1}{2} & \frac{1}{4} & 0\\ 0 & 0 & -4 & -2 & 1 & 0 \end{array}\right)⇝−14(iii)(100−1010101214000112−140)⇒TAB=(TBA)−1=(−1011214012−140)\stackrel{-\frac{1}{4} (iii)}{\leadsto} \left(\begin{array}{ccc|ccc} 1 & 0 & 0 & -1 & 0 & 1\\ 0 & 1 & 0 & \frac{1}{2} & \frac{1}{4} & 0\\ 0 & 0 & 1 & \frac{1}{2} & -\frac{1}{4} & 0 \end{array}\right) \quad\quad \Rightarrow \quad\quad T_A^B = (T_B^A)^{-1} = \begin{pmatrix} -1 & 0 & 1\\ \frac{1}{2} & \frac{1}{4} & 0\\ \frac{1}{2} & -\frac{1}{4} & 0 \end{pmatrix}

b) Using the calculation rule we obtain:

MAA(F)=TABMBB(F)TBA=(−1011214012−140)(010002000)(01102−2111)=(0−221232−12−1212−32)\begin{aligned} M_A^A(F) &= T_A^B M_B^B(F) T_B^A\\ &= \begin{pmatrix} -1 & 0 & 1\\ \frac{1}{2} & \frac{1}{4} & 0\\ \frac{1}{2} & -\frac{1}{4} & 0 \end{pmatrix} \begin{pmatrix} 0 & 1 & 0\\ 0 & 0 & 2\\ 0 & 0 & 0 \end{pmatrix} \begin{pmatrix} 0 & 1 & 1\\ 0 & 2 & -2\\ 1 & 1 & 1 \end{pmatrix} = \begin{pmatrix} 0 & -2 & 2\\ \frac{1}{2} & \frac{3}{2} & -\frac{1}{2}\\ -\frac{1}{2} & \frac{1}{2} & -\frac{3}{2} \end{pmatrix} \end{aligned}

c) Coordinates of pp with respect to B\mathcal{B}:

pB=(2−83).p_\mathcal{B} = \begin{pmatrix} 2\\ -8\\ 3 \end{pmatrix}.

Coordinates of pp with respect to A\mathcal{A}:

pA=TABpB=(1−13).p_\mathcal{A} = T_A^B p_\mathcal{B} = \begin{pmatrix} 1\\ -1\\ 3 \end{pmatrix}.

Check: a1−a2+3a3=3x2−8x+2=p(x)✓a_1 - a_2 + 3a_3 = 3x^2 - 8x + 2 = p(x) \quad\quad \checkmark

Definition

Two matrices A,B∈Em×nA,B \in \mathbb{E}^{m \times n} are called equivalent if there exist S∈Em×mS \in \mathbb{E}^{m \times m} and T∈En×nT \in \mathbb{E}^{n \times n} with:

B=SAT−1B = SAT^{-1}

If m=nm=n, we call A,B∈Em×nA,B \in \mathbb{E}^{m \times n} similar if there exists an S∈Em×mS \in \mathbb{E}^{m \times m} with:

B=SAS−1.B = SAS^{-1}.