本文为《Linear algebra and its applications》的读书笔记
Transformations
- The difference between a matrix equation A x = b A\boldsymbol x=\boldsymbol b Ax=b and the associated vector equation x 1 a 2 + . . . + x n a n = b x_1\boldsymbol a_2+...+x_n\boldsymbol a_n=\boldsymbol b x1a2+...+xnan=b is merely a matter of notation.
- However, a matrix equation
A
x
=
b
A\boldsymbol x=\boldsymbol b
Ax=b can arise in a way that is not directly connected with linear combinations of vectors. This happens when we think of the matrix
A
A
A as an object that “acts” on a vector
x
\boldsymbol x
x by multiplication to produce a new vector called
A
x
A\boldsymbol x
Ax.
- For instance, the equations
say that multiplication by A A A transforms x \boldsymbol x x into b \boldsymbol b b and transforms u \boldsymbol u u into the zero vector. See Figure 1.
- For instance, the equations
- From this new point of view, solving the equation A x = b A\boldsymbol x=\boldsymbol b Ax=b amounts to finding all vectors x \boldsymbol x x in R 4 \mathbb R^4 R4 that are transformed into the vector b \boldsymbol b b in R 2 \mathbb R^2 R2 under the “action” of multiplication by A A A
- The correspondence from x \boldsymbol x x to A x A\boldsymbol x Ax is a f u n c t i o n function function from one set of vectors to another.
- A transformation (变换) (or function (函数) or mapping (映射))
T
T
T from
R
n
\mathbb R^n
Rn to
R
m
\mathbb R^m
Rm is a rule that assigns to each vector
x
\boldsymbol x
x in
R
n
\mathbb R^n
Rn a vector
T
(
x
)
T(\boldsymbol x)
T(x) in
R
m
\mathbb R^m
Rm. The set
R
n
\mathbb R^n
Rn is called the domain (定义域) of
T
T
T , and
R
m
\mathbb R^m
Rm is called the codomain (余定义域 / 取值空间) of
T
T
T . For
x
\boldsymbol x
x in
R
n
\mathbb R^n
Rn, the vector
T
(
x
)
T(\boldsymbol x)
T(x) in
R
m
\mathbb R^m
Rm is called the image (像) of
x
\boldsymbol x
x (under the action of
T
T
T ). The set of all images
T
(
x
)
T(\boldsymbol x)
T(x) is called the range (值域) of
T
T
T . See Figure 2.
Note that the range of T T T is the set of all linear combinations of the columns of A A A
Matrix Transformations
矩阵变换
EXAMPLE 2 (投影变换)
- If
A
=
[
1
0
0
0
1
0
0
0
0
]
A=\begin{bmatrix}1&0&0\\0&1&0\\0&0&0\end{bmatrix}
A=⎣⎡100010000⎦⎤, then the transformation
x
↦
A
x
\boldsymbol x\mapsto A\boldsymbol x
x↦Ax projects points in
R
3
\mathbb R^3
R3 onto the
x
1
x
2
x_1x_2
x1x2-plane because
EXAMPLE 3
- Let A = [ 1 3 0 1 ] A=\begin{bmatrix}1&3\\0&1\end{bmatrix} A=[1031]. The transformation T : R 2 ↦ R 2 T:\mathbb R^2\mapsto \mathbb R^2 T:R2↦R2 is called a shear transformation (剪切变换). T T T deforms the square as if the top of the square were pushed to the right while the base is held fixed.
Linear Transformations
- Theorem 5 in Section 1.4 shows that if
A
A
A is
m
×
n
m \times n
m×n, then the transformation
x
→
A
x
\boldsymbol x\rightarrow A\boldsymbol x
x→Ax has the properties
These properties identify the most important class of transformations in linear algebra.
DEFINITION
- Every matrix transformation is a linear transformation. Important examples of linear transformations that are not matrix transformations will be discussed in Chapters 4 and 5.
线性相关与线性变换的联系:如果 { v 1 , v 2 , v 3 \boldsymbol v_1,\boldsymbol v_2,\boldsymbol v_3 v1,v2,v3} 线性相关, T T T 为一线性变换,则 { T ( v 1 ) , T ( v 2 ) , T ( v 3 ) T(\boldsymbol v_1),T(\boldsymbol v_2),T(\boldsymbol v_3) T(v1),T(v2),T(v3)} 线性相关
- Property (i) says that the result T ( u + v ) T(\boldsymbol u+\boldsymbol v) T(u+v) of first adding u \boldsymbol u u and v \boldsymbol v v in R n \mathbb R^n Rn and then applying T T T is the same as first applying T T T to u \boldsymbol u u and to v \boldsymbol v v and then adding T ( u ) T(\boldsymbol u) T(u) and T ( v ) T(\boldsymbol v) T(v) in R m \mathbb R^m Rm. These two properties lead easily to the following useful facts:
如果从群论角度来看,由性质 (i) 可知, T T T 实际上可以看作群 < R n , + > <\mathbb R^n,+> <Rn,+> 到群 < R m , + > <\mathbb R^m,+> <Rm,+> 的一个群同态映射,群同态映射保持幺元,从而得到了上面的性质 (3)
- Observe that if a transformation satisfies (4) for all u \boldsymbol u u, v \boldsymbol v v and c , d c,d c,d, it must be linear. (Set c = d = 1 c = d = 1 c=d=1 for preservation of addition, and set d = 0 d = 0 d=0 for preservation of scalar multiplication.)
- Repeated application of (4) produces a useful generalization:
- In engineering and physics, (5) is referred to as a s u p e r p o s i t i o n p r i n c i p l e superposition\ principle superposition principle (叠加原理). Think of v 1 , . . . , v p \boldsymbol v_1,...,\boldsymbol v_p v1,...,vp as signals that go into a system and T ( v 1 ) , . . . , T ( v p ) T(\boldsymbol v_1),...,T(\boldsymbol v_p) T(v1),...,T(vp) as the responses of that system to the signals. The system satisfies the superposition principle if whenever an input is expressed as a linear combination of such signals, the system’s response is the same linear combination of the responses to the individual signals.
Affine transformation (仿射变换)
- An affine transformation T : R n → R m T:\mathbb R^n \rightarrow \mathbb R^m T:Rn→Rm has the form T ( x ) = A x + b T(\boldsymbol x)=A\boldsymbol x+\boldsymbol b T(x)=Ax+b, with A A A an m × n m \times n m×n matrix and b \boldsymbol b b in R m \mathbb R^m Rm. T T T is not a linear transformation when b ≠ 0 \boldsymbol b \neq \boldsymbol 0 b=0.
EXAMPLE 4
- A company manufactures two products,
B
B
B and
C
C
C. We construct a “unit cost” matrix, whose columns describe the “costs per dollar of output” for the products:
- Let
x
=
(
x
1
,
x
2
)
\boldsymbol x = (x_1,x_2)
x=(x1,x2) be a “production” vector, corresponding to
x
1
x_1
x1 dollars of product B and
x
2
x_2
x2 dollars of product C, and define
T
:
R
2
→
R
3
T:\mathbb R^2\rightarrow \mathbb R^3
T:R2→R3 by
- The mapping
T
T
T transforms a list of production quantities (measured in dollars) into a list of total costs. The linearity of this mapping is reflected in two ways:
- If production is increased by a factor of, say, 4, from x x x to 4 x 4x 4x, then the costs will increase by the same factor, from T ( x ) T(\boldsymbol x) T(x) to 4 T ( x ) 4T(\boldsymbol x) 4T(x).
- If x \boldsymbol x x and y \boldsymbol y y are production vectors, then the total cost vector associated with the combined production x + y \boldsymbol x+\boldsymbol y x+y is precisely the sum of the cost vectors T ( x ) T(\boldsymbol x) T(x) and T ( y ) T(\boldsymbol y) T(y).