本文为《Linear algebra and its applications》的读书笔记
Change of basis
- When a basis B \mathcal B B is chosen for an n n n-dimensional vector space V V V, the associated coordinate mapping onto R n \mathbb R^n Rn provides a coordinate system for V V V. Each x \boldsymbol x x in V V V is identified uniquely by its B \mathcal B B-coordinate vector [ x ] B [\boldsymbol x]_{\mathcal B} [x]B.
- In this section, we study how [ x ] C [\boldsymbol x]_{\mathcal C} [x]C and [ x ] B [\boldsymbol x]_{\mathcal B} [x]B are related for each x \boldsymbol x x in V V V.
EXAMPLE 1
Consider two bases
B
=
{
b
1
,
b
2
}
\mathcal B=\{\boldsymbol b_1,\boldsymbol b_2\}
B={b1,b2} and
C
=
{
c
1
,
c
2
}
\mathcal C=\{\boldsymbol c_1,\boldsymbol c_2\}
C={c1,c2} for a vector space
V
V
V, such that
Suppose
[
x
]
B
=
[
3
1
]
[\boldsymbol x]_{\mathcal B}=\begin{bmatrix} 3\\1\end{bmatrix}
[x]B=[31]. Find
[
x
]
C
[\boldsymbol x]_{\mathcal C}
[x]C.
SOLUTION
- Apply the coordinate mapping determined by
C
\mathcal C
C to
x
\boldsymbol x
x. Since the coordinate mapping is a linear transformation,
- Thus
[ x ] C = [ 4 − 6 1 1 ] [ 3 1 ] = [ 6 4 ] [\mathbf{x}]_{\mathcal{C}}=\left[\begin{array}{rr} 4 & -6 \\ 1 & 1 \end{array}\right]\left[\begin{array}{l} 3 \\ 1 \end{array}\right]=\left[\begin{array}{l} 6 \\ 4 \end{array}\right] [x]C=[41−61][31]=[64]
The theorem shows that to solve the change-of-basis problem, we need the coordinate vectors of the old basis relative to the new basis.
- The matrix
P
C
←
B
\mathop{P}\limits_{\mathcal C \leftarrow \mathcal B}
C←BP is called the change-of-coordinates matrix from
B
\mathcal B
B to
C
\mathcal C
C (坐标变换矩阵). Multiplication by
P
C
←
B
\mathop{P}\limits_{\mathcal C \leftarrow \mathcal B}
C←BP converts
B
\mathcal B
B-coordinates into
C
\mathcal C
C-coordinates. Figure 2 illustrates the change-of-coordinates equation (4).
- Since
P
C
←
B
\mathop{P}\limits_{\mathcal C \leftarrow \mathcal B}
C←BP is invertible,
( P C ← B ) − 1 [ x ] C = [ x ] B (\mathop{P}\limits_{\mathcal C \leftarrow \mathcal B})^{-1}[\boldsymbol x]_{\mathcal C}=[\boldsymbol x]_{\mathcal B} (C←BP)−1[x]C=[x]BThus ( P C ← B ) − 1 (\mathop{P}\limits_{\mathcal C \leftarrow \mathcal B})^{-1} (C←BP)−1 is the matrix that converts C \mathcal C C-coordinates into B \mathcal B B-coordinates. That is,
( P C ← B ) − 1 = P B ← C (\mathop{P}\limits_{\mathcal C \leftarrow \mathcal B})^{-1}=\mathop{P}\limits_{\mathcal B \leftarrow \mathcal C} (C←BP)−1=B←CP
EXAMPLE 2
Let b 1 = [ − 9 1 ] , b 2 = [ − 5 − 1 ] , c 1 = [ 1 − 4 ] , c 2 = [ 3 − 5 ] \boldsymbol b_1=\begin{bmatrix} -9\\1\end{bmatrix},\boldsymbol b_2=\begin{bmatrix} -5\\-1\end{bmatrix},\boldsymbol c_1=\begin{bmatrix} 1\\-4\end{bmatrix},\boldsymbol c_2=\begin{bmatrix} 3\\-5\end{bmatrix} b1=[−91],b2=[−5−1],c1=[1−4],c2=[3−5], and consider the bases for R 2 \mathbb R^2 R2 given by B = { b 1 , b 2 } \mathcal B = \{\boldsymbol b_1,\boldsymbol b_2\} B={b1,b2} and C = { c 1 , c 2 } \mathcal C = \{\boldsymbol c_1,\boldsymbol c_2\} C={c1,c2}. Find the change-of-coordinates matrix from B \mathcal B B to C \mathcal C C.
SOLUTION
- To solve both systems simultaneously, augment the coefficient matrix with
b
1
\boldsymbol b_1
b1 and
b
2
\boldsymbol b_2
b2, and row reduce:
- Observe that
In general,
Change of Basis in R n \mathbb R^n Rn
- If
B
=
{
b
1
,
.
.
.
,
b
n
}
\mathcal B=\{\boldsymbol b_1,...,\boldsymbol b_n\}
B={b1,...,bn} and
ε
ε
ε is the standard basis
{
e
1
,
.
.
.
,
e
n
}
\{\boldsymbol e_1,...,\boldsymbol e_n\}
{e1,...,en} in
R
n
\mathbb R^n
Rn, then
[
b
1
]
ε
=
b
1
[\boldsymbol b_1]_ε=\boldsymbol b_1
[b1]ε=b1, and likewise for the other vectors in
B
\mathcal B
B. In this case,
P
ε
←
B
\mathop{P}\limits_{ ε \leftarrow \mathcal B}
ε←BP is the same as the change-of-coordinates matrix
P
B
P_{\mathcal B}
PB introduced in Section 4.4, namely,
P B = [ b 1 . . . b n ] P_{\mathcal B}=[\boldsymbol b_1\ \ ...\ \ \boldsymbol b_n] PB=[b1 ... bn] - Recall that for each
x
\boldsymbol x
x in
R
n
\mathbb R^n
Rn,
P C [ x ] C = x a n d P B [ x ] B = x P_{\mathcal C}[\boldsymbol x]_{\mathcal C}=\boldsymbol x\ \ \ and\ \ \ P_{\mathcal B}[\boldsymbol x]_{\mathcal B}=\boldsymbol x PC[x]C=x and PB[x]B=xThus
[ x ] C = P C − 1 x = P C − 1 P B [ x ] B [\boldsymbol x]_{\mathcal C}=P_{\mathcal C}^{-1}\boldsymbol x=P_{\mathcal C}^{-1}P_{\mathcal B}[\boldsymbol x]_{\mathcal B} [x]C=PC−1x=PC−1PB[x]BIn R n \mathbb R^n Rn, the change-of-coordinates matrix P C ← B \mathop{P}\limits_{\mathcal C \leftarrow \mathcal B} C←BP may be computed as P C − 1 P B P_{\mathcal C}^{-1}P_{\mathcal B} PC−1PB.