本文为《Linear algebra and its applications》的读书笔记
目录
The Inner Product
- The number
u
T
v
\boldsymbol u^T \boldsymbol v
uTv is called the inner product of
u
\boldsymbol u
u and
v
\boldsymbol v
v, and often it is written as
u
⋅
v
\boldsymbol u\cdot \boldsymbol v
u⋅v. This inner product is also referred to as a dot product (点积).
- Properties (
b
b
b) and (
c
c
c) can be combined several times to produce the following useful rule:
The Length of a Vector
- For any scalar
c
c
c, the length of
c
v
c\boldsymbol v
cv is
∣
c
∣
|c|
∣c∣ times the length of
v
\boldsymbol v
v. That is,
- A vector whose length is 1 1 1 is called a unit vector. If we divide a nonzero vector v \boldsymbol v v by its length—that is, multiply by 1 / ∥ v ∥ 1/\left \| \boldsymbol v\right \| 1/∥v∥—we obtain a unit vector u \boldsymbol u u. This process is sometimes called normalizing (单位化) v \boldsymbol v v.
Distance in R n \mathbb R^n Rn
- Recall that if a a a and b b b are real numbers, the distance on the number line between a a a and b b b is the number ∣ a − b ∣ |a - b| ∣a−b∣. This definition of distance in R \mathbb R R has a direct analogue in R n \mathbb R^n Rn.
- In R 2 \mathbb R^2 R2 and R 3 \mathbb R^3 R3, this definition of distance coincides with the usual formulas for the Euclidean distance between two points.
Orthogonal Vectors
正交向量
- The rest of this chapter depends on the fact that the concept of perpendicular lines in ordinary Euclidean geometry has an analogue in R n \mathbb R^n Rn.
- Consider
R
2
\mathbb R^2
R2 or
R
3
\mathbb R^3
R3 and two lines through the origin determined by vectors
u
\boldsymbol u
u and
v
\boldsymbol v
v. The two lines shown in Figure 5 are geometrically perpendicular if and only if the distance from
u
\boldsymbol u
u to
v
\boldsymbol v
v is the same as the distance from
u
\boldsymbol u
u to
−
v
-\boldsymbol v
−v.
The two squared distances are equal if and only if u ⋅ v = 0 \boldsymbol u\cdot \boldsymbol v=0 u⋅v=0.
- The following definition generalizes to
R
n
\mathbb R^n
Rn this notion of perpendicularity (or
o
r
t
h
o
g
o
n
a
l
i
t
y
\boldsymbol{orthogonality}
orthogonality).
Observe that the zero vector is orthogonal to every vector in R n \mathbb R^n Rn.
毕达哥拉斯定理 / 勾股定理
Orthogonal Complements
正交补
- If a vector z \boldsymbol z z is orthogonal to every vector in a subspace W \mathbb W W of R n \mathbb R^n Rn, then z \boldsymbol z z is said to be orthogonal to W \mathbb W W . The set of all vectors z \boldsymbol z z that are orthogonal to W \mathbb W W is called the orthogonal complement of W \mathbb W W and is denoted by W ⊥ \mathbb W^\perp W⊥ (and read as “ W \mathbb W W perpendicular” or simply “ W \mathbb W W perp”).
( R o w A ) ⊥ = N u l A ; ( C o l A ) ⊥ = N u l A T (RowA)^\perp=NulA;(ColA)^\perp=NulA^T (RowA)⊥=NulA;(ColA)⊥=NulAT
In fact, Theorem 3 is sometimes called the Fundamental Theorem of Linear Algebra.
PROOF
- The row–column rule for computing A x A\boldsymbol x Ax shows that if x \boldsymbol x x is in N u l A NulA NulA, then x \boldsymbol x x is orthogonal to each row of A A A. Since the rows of A A A span the row space, x \boldsymbol x x is orthogonal to R o w A RowA RowA. Conversely, if x \boldsymbol x x is orthogonal to R o w A RowA RowA, then A x = 0 A\boldsymbol x=\boldsymbol 0 Ax=0.
- This proves the first statement of the theorem. Since this statement is true for any matrix, it is true for
A
T
A^T
AT . That is, the orthogonal complement of the row space of
A
T
A^T
AT is the null space of
A
T
A^T
AT . This proves the second statement, because
R
o
w
A
T
=
C
o
l
A
RowA^T = ColA
RowAT=ColA.
d i m W + d i m W ⊥ = n dim W +dim W^\perp = n dimW+dimW⊥=n
EXAMPLE
Let W W W be a subspace of R n \mathbb R^n Rn. Thus W ⊥ W^\perp W⊥ is also a subspace of R n \mathbb R^n Rn. Prove that d i m W + d i m W ⊥ = n dim W +dim W^\perp = n dimW+dimW⊥=n.
SOLUTION
- If W ≠ { 0 } W \neq \{\boldsymbol 0\} W={0}, let { b 1 , . . . , b p } \{\boldsymbol b_1,..., \boldsymbol b_p\} {b1,...,bp} be a basis for W W W. Let A A A be the p × n p \times n p×n matrix having rows b 1 T , . . . , b p T \boldsymbol b_1^T,...,\boldsymbol b_p^T b1T,...,bpT. It follows that W W W is the row space of A A A. Theorem 3 implies that W ⊥ = ( R o w A ) ⊥ = N u l A W^\perp= (Row A)^\perp= Nul A W⊥=(RowA)⊥=NulA and hence d i m W ⊥ = d i m N u l A dim W ^\perp= dim Nul A dimW⊥=dimNulA. Thus, d i m W + d i m W ⊥ = d i m R o w A + d i m N u l A = n dim W + dimW^\perp = dim Row A+ dim Nul A = n dimW+dimW⊥=dimRowA+dimNulA=n.
- If W = { 0 } W =\{\boldsymbol 0\} W={0}, then W ⊥ = R n W^\perp =\mathbb R^n W⊥=Rn, and the result follows.
EXERCISES
Explain why an equation A x = b A\boldsymbol x = \boldsymbol b Ax=b has a solution if and only if b \boldsymbol b b is orthogonal to all solutions of the equation A T x = 0 A^T\boldsymbol x = \boldsymbol 0 ATx=0.
SOLUTION
- If A x = b A\boldsymbol x = \boldsymbol b Ax=b has a solution, then x T A T = b T . \boldsymbol x^TA^T=\boldsymbol b^T. xTAT=bT. Let x ′ \boldsymbol x' x′ be any solution of A T x = 0 A^T\boldsymbol x = \boldsymbol 0 ATx=0, then x T A T x ′ = b T x ′ = 0 , b ⋅ x ′ = 0 \boldsymbol x^TA^T\boldsymbol x'=\boldsymbol b^T\boldsymbol x'=0,\boldsymbol b\cdot\boldsymbol x'=0 xTATx′=bTx′=0,b⋅x′=0. Thus b \boldsymbol b b is orthogonal to all solutions of the equation A T x = 0 A^T\boldsymbol x = \boldsymbol 0 ATx=0.
- If b \boldsymbol b b is orthogonal to all solutions of the equation A T x = 0 A^T\boldsymbol x = \boldsymbol 0 ATx=0, then b ∈ ( N u l A T ) ⊥ = C o l A \boldsymbol b\in(NulA^T)^\perp=ColA b∈(NulAT)⊥=ColA. Thus A x = b A\boldsymbol x = \boldsymbol b Ax=b has a solution.
Angles in R 2 \mathbb R^2 R2 and R 3 \mathbb R^3 R3
- If
u
\boldsymbol u
u and
v
\boldsymbol v
v are nonzero vectors in either
R
2
\mathbb R^2
R2 or
R
3
\mathbb R^3
R3, then there is a nice connection between their inner product and the angle
θ
\theta
θ between the two line segments from the origin to the points identified with
u
\boldsymbol u
u and
v
\boldsymbol v
v. The formula is
- To verify this formula for vectors in
R
2
\mathbb R^2
R2, consider the triangle shown in Figure 9.
By the law of cosines (余弦定理),
which can be rearranged to produce
The verification for R 3 \mathbb R^3 R3 is similar. When n > 3 n > 3 n>3, this formula may be used to define the angle between two vectors in R n \mathbb R^n Rn. - In statistics, for instance, the value of c o s θ cos \theta cosθ for suitable vectors u \boldsymbol u u and v \boldsymbol v v is what statisticians call a correlation coefficient (相关系数).