《矩阵理论》大萌课程笔记 - 线性变换
总目录
章节名称与链接 | |
---|---|
线性空间与线性变换 | 线性空间与子空间 |
有限维线性空间、基、维数 | |
线性变换 | |
内积空间 | |
特征值与特征向量 | |
特殊矩阵 | |
矩阵分解 | |
矩阵函数 |
声明
本专栏博客用于记录上海交通大学研究生课程《矩阵理论》笔记,课程任教老师为邓大萌老师。所有内容均为博主个人的课堂笔记,包括课堂例题与证明。如有不妥、错误之处欢迎大家指正。
文章目录
1 线性变换定义与性质
1.1 定义
U , V U,V U,V是 F F F上的线性空间, σ : U → V \sigma:U\rightarrow V σ:U→V表示 U U U到 V V V的映射, ∀ α , β ∈ U , k ∈ F \forall \alpha,\beta\in U,k\in F ∀α,β∈U,k∈F,满足以下条件:
1) σ ( α + β ) = σ ( α ) + σ ( β ) \sigma(\alpha+\beta)=\sigma(\alpha)+\sigma(\beta) σ(α+β)=σ(α)+σ(β)
2) σ ( k ⋅ α ) = k ⋅ σ ( α ) \sigma(k\cdot \alpha)=k\cdot\sigma(\alpha) σ(k⋅α)=k⋅σ(α)
则称 σ \sigma σ是 U U U到 V V V的线性变换。
例1: U → 0 U\rightarrow 0 U→0:零变换
例2: U → U U\rightarrow U U→U:恒等变换
例3: ∀ f ∈ P [ R ] , σ ( f ) = f ′ : P [ R ] → P [ R ] \forall f\in P[R], \sigma(f)=f' : P[R]\rightarrow P[R] ∀f∈P[R],σ(f)=f′:P[R]→P[R]:多项式求导变换
例4: ∀ f ∈ P [ R ] , σ ( f ) = ∫ 0 1 f ( x ) d x : P [ R ] → P [ R ] \forall f\in P[R], \sigma(f)=\int_0^1f(x)dx : P[R]\rightarrow P[R] ∀f∈P[R],σ(f)=∫01f(x)dx:P[R]→P[R]:多项式积分变换
例5: ∀ f ∈ P [ R ] , σ ( f ) = x 2 f : P [ R ] → P [ R ] \forall f\in P[R], \sigma(f)=x^2f : P[R]\rightarrow P[R] ∀f∈P[R],σ(f)=x2f:P[R]→P[R]
例6: ∀ x ∈ R , σ ( x ) = a x ( a > 0 , a ≠ 1 , a 固 定 ) : R → R + \forall x\in R,\sigma(x)=a^x(a>0,a\ne1,a固定) : R\rightarrow R^+ ∀x∈R,σ(x)=ax(a>0,a=1,a固定):R→R+
⋆ ⋆ ⋆ \star\star\star ⋆⋆⋆
∀ α , β ∈ R , k ∈ R \forall \alpha,\beta\in R,k\in R ∀α,β∈R,k∈R
σ ( α + β ) = a α + β = a α ⋅ a β = σ ( α ) ⊕ σ ( β ) \sigma(\alpha+\beta)=a^{\alpha+\beta}=a^\alpha\cdot a^\beta=\sigma(\alpha)\oplus\sigma(\beta) σ(α+β)=aα+β=aα⋅aβ=σ(α)⊕σ(β)
σ ( k ⋅ α ) = a k ⋅ α = ( a α ) k = k ⊗ σ ( α ) \sigma(k\cdot \alpha)=a^{k\cdot\alpha}=(a^\alpha)^k=k\otimes\sigma(\alpha) σ(k⋅α)=ak⋅α=(aα)k=k⊗σ(α)
例7:举出满足第一条不满足第二条、满足第二条不满足第一条、两条都不满足的示例
满足第一条不满足第二条(复数): σ : C → C , σ ( a + b i ) = b \sigma :C\rightarrow C,\sigma(a+bi)=b σ:C→C,σ(a+bi)=b
验证: σ ( ( a 1 + b 1 i ) + ( a 2 + b 2 i ) ) = σ ( ( a 1 + a 2 ) + ( b 1 + b 2 ) i ) = b 1 + b 2 \sigma((a_1+b_1i)+(a_2+b_2i))=\sigma((a_1+a_2)+(b_1+b_2)i)=b_1+b_2 σ((a1+b1i)+(a2+b2i))=σ((a1+a2)+(b1+b2)i)=b1+b2
σ ( k ⋅ ( a 1 + b 1 i ) ) \sigma(k\cdot(a_1+b_1i)) σ(k⋅(a1+b1i)),当 k = i k=i k=i时不满足封闭性
满足第二条不满足第一条(范数): σ : R 2 → R , σ ( x , y ) = x 2 + y 2 \sigma:R^2\rightarrow R,\sigma(x,y)=\sqrt{x^2+y^2} σ:R2→R,σ(x,y)=x2+y2
两条都不满足的示例(非线性函数): σ : R → R , σ ( x ) = 1 x \sigma:R\rightarrow R,\sigma(x)=\frac{1}{x} σ:R→R,σ(x)=x1
1.2 性质
(1) 0 → 0 0\rightarrow 0 0→0
(2) σ ( − α ) = − σ ( α ) \sigma(-\alpha)=-\sigma(\alpha) σ(−α)=−σ(α)
(3) 若 α 1 , α 2 , . . . , α n \alpha_1,\alpha_2,...,\alpha_n α1,α2,...,αn线性相关,则 σ ( α 1 , α 2 , . . . , α n ) \sigma(\alpha_1,\alpha_2,...,\alpha_n) σ(α1,α2,...,αn)也线性相关;逆命题不成立,逆否命题成立。
(4) α 1 , α 2 , . . . , α n \alpha_1,\alpha_2,...,\alpha_n α1,α2,...,αn 为 U U U的一组基, β 1 , β 2 , . . . , β n \beta_1,\beta_2,...,\beta_n β1,β2,...,βn为 V V V的一组基,则存在唯一一个线性变换满足 σ ( α i ) = β i \sigma(\alpha_i)=\beta_i σ(αi)=βi
等价于:线性变换由基的像唯一确定。
2 线性变换空间
2.1 定义
U到V所有线性变换的集合所构成的空间称为线性变换空间,用 L ( U , V ) L(U,V) L(U,V)表示
作业1:验证 L ( U , V ) L(U,V) L(U,V)是线性空间
解:设 σ 1 , σ 2 \sigma_1,\sigma_2 σ1,σ2是 L ( U , V ) L(U,V) L(U,V)的两个线性变换, α 1 , α 2 ∈ U , k , K ∈ F \alpha_1,\alpha_2\in U, k,K\in F α1,α2∈U,k,K∈F
则 ( σ 1 + σ 2 ) ( α 1 + α 2 ) = σ 1 ( α 1 + α 2 ) + σ 2 ( α 1 + α 2 ) = σ 1 ( α 1 ) + σ 1 ( α 2 ) + σ 2 ( α 1 ) + σ 2 ( α 2 ) = ( σ 1 + σ 2 ) ( α 1 ) + ( σ 1 + σ 2 ) ( α 2 ) (\sigma_1+\sigma_2)(\alpha_1+\alpha_2)=\sigma_1(\alpha_1+\alpha_2)+\sigma_2(\alpha_1+\alpha_2)=\sigma_1(\alpha_1)+\sigma_1(\alpha_2)+\sigma_2(\alpha_1)+\sigma_2(\alpha_2)=(\sigma_1+\sigma_2)(\alpha_1)+(\sigma_1+\sigma_2)(\alpha_2) (σ1+σ2)(α1+α2)=σ1(α1+α2)+σ2(α1+α2)=σ1(α1)+σ1(α2)+σ2(α1)+σ2(α2)=(σ1+σ2)(α1)+(σ1+σ2)(α2)
( σ 1 + σ 2 ) ( k ⋅ α 1 ) = σ 1 ( k ⋅ α 1 ) + σ 2 ( k ⋅ α 1 ) = k ⋅ σ 1 ( α 1 ) + k ⋅ σ 2 ( α 1 ) = k ⋅ ( σ 1 + σ 2 ) ( α 1 ) (\sigma_1+\sigma_2)(k\cdot \alpha_1)=\sigma_1(k\cdot \alpha_1)+\sigma_2(k\cdot \alpha_1)=k\cdot \sigma_1(\alpha_1)+k\cdot \sigma_2(\alpha_1)=k\cdot (\sigma_1+\sigma_2)(\alpha_1) (σ1+σ2)(k⋅α1)=σ1(k⋅α1)+σ2(k⋅α1)=k⋅σ1(α1)+k⋅σ2(α1)=k⋅(σ1+σ2)(α1)
⇒ σ 1 + σ 2 ∈ L ( U , V ) \Rightarrow \sigma_1+\sigma_2\in L(U,V) ⇒σ1+σ2∈L(U,V)
又
( K ⋅ σ 1 ) ( α 1 + α 2 ) = K ⋅ σ 1 ( α 1 + α 2 ) = K ⋅ σ 1 ( α 1 ) + K ⋅ σ 1 ( α 2 ) = K ⋅ σ 1 ( α 1 + α 2 ) (K\cdot \sigma_1)(\alpha_1+\alpha_2)=K\cdot \sigma_1(\alpha_1+\alpha_2)=K\cdot \sigma_1(\alpha_1)+K\cdot \sigma_1(\alpha_2)=K\cdot \sigma_1(\alpha_1+\alpha_2) (K⋅σ1)(α1+α2)=K⋅σ1(α1+α2)=K⋅σ1(α1)+K⋅σ1(α2)=K⋅σ1(α1+α2)
( K ⋅ σ 1 ) ( k ⋅ α 1 ) = K ⋅ σ 1 ( k ⋅ α 1 ) = K ⋅ k ⋅ σ 1 ( α 1 ) = k ⋅ ( K ⋅ σ 1 ) ( α 1 ) (K\cdot \sigma_1)(k\cdot \alpha_1)=K\cdot\sigma_1(k\cdot\alpha_1)=K\cdot k\cdot\sigma_1(\alpha_1)=k\cdot(K\cdot\sigma_1)(\alpha_1) (K⋅σ1)(k⋅α1)=K⋅σ1(k⋅α1)=K⋅k⋅σ1(α1)=k⋅(K⋅σ1)(α1)
⇒ K ⋅ σ 1 ∈ L ( U , V ) \Rightarrow K\cdot\sigma_1\in L(U,V) ⇒K⋅σ1∈L(U,V)
∴ L ( U , V ) \therefore L(U,V) ∴L(U,V)满足计算封闭性
∴ L ( U , V ) \therefore L(U,V) ∴L(U,V)是线性空间。
2.2 复合线性变换空间
若 T 1 ∈ L ( U , V ) , T 2 ∈ L ( V , W ) T_1\in L(U,V),T_2\in L(V, W) T1∈L(U,V),T2∈L(V,W),则 T 1 T_1 T1与 T 2 T_2 T2的复合变换表示为 : T 1 ⋅ T 2 = T 1 ( T 2 ( U ) ) T_1\cdot T_2=T_1(T_2(U)) T1⋅T2=T1(T2(U))
作业2:验证 T 1 ⋅ T 2 T_1\cdot T_2 T1⋅T2是线性变换:
设 α 1 , α 2 ∈ U , k ∈ F \alpha_1,\alpha_2\in U, k\in F α1,α2∈U,k∈F
则: ( T 1 ⋅ T 2 ) ( α 1 + α 2 ) = T 1 ( T 2 ( α 1 + α 2 ) ) = T 1 ( T 2 ( α 1 ) + T 2 ( α 2 ) ) = ( T 1 ⋅ T 2 ) ( α 1 ) + ( T 1 ⋅ T 2 ) ( α 2 ) (T_1\cdot T_2)(\alpha_1+\alpha_2)=T_1(T_2(\alpha_1+\alpha_2))=T_1(T_2(\alpha_1)+T_2(\alpha_2))=(T_1\cdot T_2)(\alpha_1)+(T_1\cdot T_2)(\alpha_2) (T1⋅T2)(α1+α2)=T1(T2(α1+α2))=T1(T2(α1)+T2(α2))=(T1⋅T2)(α1)+(T1⋅T2)(α2)
( T 1 ⋅ T 2 ) ( k ⋅ α 1 ) = T 1 ( T 2 ( k ⋅ α 1 ) ) = T 1 ( k ⋅ T 2 ( α 1 ) ) = k ⋅ ( T 1 ⋅ T 2 ) ( α 1 ) (T_1\cdot T_2)(k\cdot \alpha_1)=T_1(T_2(k\cdot \alpha_1))=T_1(k\cdot T_2(\alpha_1))=k\cdot (T_1\cdot T_2)(\alpha_1) (T1⋅T2)(k⋅α1)=T1(T2(k⋅α1))=T1(k⋅T2(α1))=k⋅(T1⋅T2)(α1)
⇒ T 1 ⋅ T 2 \Rightarrow T_1\cdot T_2 ⇒T1⋅T2是线性变换
复合变换性质:
- ( T 1 ⋅ T 2 ) ⋅ T 3 = T 1 ⋅ ( T 2 ⋅ T 3 ) (T_1\cdot T_2)\cdot T_3=T_1\cdot(T_2\cdot T_3) (T1⋅T2)⋅T3=T1⋅(T2⋅T3)
- T 1 ⋅ ( T 2 + T 3 ) = T 1 ⋅ T 2 + T 1 ⋅ T 3 T_1\cdot (T_2+T_3)=T_1\cdot T_2+T_1\cdot T_3 T1⋅(T2+T3)=T1⋅T2+T1⋅T3
- ( T 1 + T 2 ) ⋅ T 3 = T 1 ⋅ T 3 + T 2 ⋅ T 3 (T_1+T_2)\cdot T3=T_1\cdot T_3+T_2\cdot T_3 (T1+T2)⋅T3=T1⋅T3+T2⋅T3
证明两条分配律,即验证:
- T 1 ⋅ ( T 2 + T 3 ) = T 1 ⋅ T 2 + T 1 ⋅ T 3 , T 1 ∈ L ( U , V ) , T 2 , T 3 ∈ L ( V , W ) T_1\cdot (T_2+T_3)=T_1\cdot T_2+T_1\cdot T_3, T_1\in L(U,V), T_2,T_3\in L(V,W) T1⋅(T2+T3)=T1⋅T2+T1⋅T3,T1∈L(U,V),T2,T3∈L(V,W)
- ( T 1 + T 2 ) ⋅ T 3 = T 1 ⋅ T 3 + T 2 ⋅ T 3 , T 1 , T 2 ∈ L ( U , V ) , T 3 ∈ L ( V , W ) (T_1+T_2)\cdot T3=T_1\cdot T_3+T_2\cdot T_3, T_1,T_2\in L(U,V), T_3\in L(V,W) (T1+T2)⋅T3=T1⋅T3+T2⋅T3,T1,T2∈L(U,V),T3∈L(V,W)
证明:
设 α ∈ U \alpha \in U α∈U
则: ( T 1 ⋅ ( T 2 + T 3 ) ) ( α ) = T 1 ⋅ ( T 2 + T 3 ) ( α ) = T 1 ⋅ ( T 2 ( α ) + T 3 ( α ) ) = T 1 ⋅ ( T 2 ( α ) ) + T 1 ⋅ ( T 3 ( α ) ) = ( T 1 ⋅ T 2 ) ( α ) + ( T 1 ⋅ T 3 ) ( α ) (T_1\cdot(T_2+T_3))(\alpha)=T_1\cdot(T_2+T_3)(\alpha)=T_1\cdot(T_2(\alpha)+T_3(\alpha))=T_1\cdot (T_2(\alpha))+T_1\cdot(T_3(\alpha))=(T_1\cdot T_2)(\alpha)+(T_1\cdot T_3)(\alpha) (T1⋅(T2+T3))(α)=T1⋅(T2+T3)(α)=T1⋅(T2(α)+T3(α))=T1⋅(T2(α))+T1⋅(T3(α))=(T1⋅T2)(α)+(T1⋅T3)(α)
( ( T 1 + T 2 ) ⋅ T 3 ) ( α ) = ( T 1 + T 2 ) ⋅ ( T 3 ( α ) ) = T 1 ( T 3 ( α ) ) + T 2 ( T 3 ( α ) ) = ( T 1 ⋅ T 3 ) ( α ) + ( T 2 ⋅ T 3 ) ( α ) ((T_1+T_2)\cdot T_3)(\alpha)=(T_1+ T_2)\cdot (T_3(\alpha))=T_1(T_3(\alpha))+T_2(T_3(\alpha))=(T_1\cdot T_3)(\alpha)+(T_2\cdot T_3)(\alpha) ((T1+T2)⋅T3)(α)=(T1+T2)⋅(T3(α))=T1(T3(α))+T2(T3(α))=(T1⋅T3)(α)+(T2⋅T3)(α)
证毕。
2.3 特殊线性变换空间
2.3.1 零空间
定义:设 σ : U → V \sigma:U\rightarrow V σ:U→V ,则 n u l l σ = { α ∈ U , σ ( α ) = 0 } null\ \sigma=\{\alpha\in U,\sigma(\alpha)=0\} null σ={α∈U,σ(α)=0}为零空间, n u l l σ ∈ U null\ \sigma\in U null σ∈U
功能:用于判别单射(单射的含义: 若 存 在 α , β 使 得 σ ( α ) = σ ( β ) , 则 α = β 若存在\alpha,\beta使得\sigma(\alpha)=\sigma(\beta),则\alpha=\beta 若存在α,β使得σ(α)=σ(β),则α=β)
证明 n u l l σ = { 0 } ⇔ 单 射 null\ \sigma=\{0\}\Leftrightarrow 单射 null σ={0}⇔单射
证明:
- 单 射 ⇒ n u l l σ = { 0 } 单射\Rightarrow null\ \sigma=\{0\} 单射⇒null σ={0}
单射时0有且只有映射到0,不能存在多个 α \alpha α
- n u l l σ = { 0 } ⇒ 单 射 null\ \sigma=\{0\}\Rightarrow 单射 null σ={0}⇒单射
若存在 α , β \alpha,\beta α,β使得 σ ( α ) = σ ( β ) ⇒ σ ( α ) − σ ( β ) = σ ( α − β ) = 0 ⇒ σ ( α − β ) ∈ n u l l σ ⇒ α − β = 0 ⇒ α = β \sigma(\alpha)=\sigma(\beta)\Rightarrow \sigma(\alpha)-\sigma(\beta)=\sigma(\alpha-\beta)=0\Rightarrow \sigma(\alpha-\beta)\in null\ \sigma\Rightarrow \alpha-\beta=0\Rightarrow \alpha=\beta σ(α)=σ(β)⇒σ(α)−σ(β)=σ(α−β)=0⇒σ(α−β)∈null σ⇒α−β=0⇒α=β
证毕。
例1: σ : U → V , α ∈ U , σ ( α ) = 0 \sigma:U\rightarrow V, \alpha\in U,\sigma(\alpha)=0 σ:U→V,α∈U,σ(α)=0(零变换): n u l l σ = U null\ \sigma=U null σ=U
例2: σ : R n × n → R n × n , A ∈ R n × n , σ ( A ) = A − A T \sigma:R^{n\times n}\rightarrow R^{n\times n}, A\in R^{n\times n},\sigma(A)=A-A^T σ:Rn×n→Rn×n,A∈Rn×n,σ(A)=A−AT: n u l l σ = { 对 称 阵 } null\ \sigma=\{对称阵\} null σ={对称阵}
例3: σ : P [ R ] → P [ R ] , f ∈ P [ R ] , σ ( f ) = f ′ \sigma : P[R]\rightarrow P[R],f \in P[R],\sigma(f)=f' σ:P[R]→P[R],f∈P[R],σ(f)=f′: n u l l σ = 常 数 ( 或 者 s p a n ( 1 ) ) null\ \sigma=常数(或者span(1)) null σ=常数(或者span(1))
例4: σ : P [ R ] → P [ R ] , f ∈ P [ R ] , σ ( f ) = x 3 f \sigma : P[R]\rightarrow P[R],f \in P[R],\sigma(f)=x^3f σ:P[R]→P[R],f∈P[R],σ(f)=x3f: n u l l σ = { 0 } null\ \sigma=\{0\} null σ={0}
2.3.2 像空间
定义:设 σ : U → V \sigma:U\rightarrow V σ:U→V ,则 r a n g e σ = { α ∈ U , σ ( α ) } range\ \sigma=\{\alpha\in U,\sigma(\alpha)\} range σ={α∈U,σ(α)}为像空间, r a n g e σ ∈ V range\ \sigma\in V range σ∈V
功能:用于判别满射(满射的含义: ∀ β ∈ V , ∃ α ∈ U , σ ( α ) = β \forall \beta\in V,\exist \alpha\in U,\sigma(\alpha)=\beta ∀β∈V,∃α∈U,σ(α)=β)
r a n g e σ = V ⟺ range\ \sigma=V\iff range σ=V⟺满射
例1: σ : U → V , α ∈ U , σ ( α ) = 0 \sigma:U\rightarrow V, \alpha\in U,\sigma(\alpha)=0 σ:U→V,α∈U,σ(α)=0(零变换): r a n g e σ = 0 range\ \sigma=0 range σ=0
例2: σ : R n × n → R n × n , A ∈ R n × n , σ ( A ) = A − A T \sigma:R^{n\times n}\rightarrow R^{n\times n}, A\in R^{n\times n},\sigma(A)=A-A^T σ:Rn×n→Rn×n,A∈Rn×n,σ(A)=A−AT: r a n g e σ = { 反 对 称 阵 } range\ \sigma=\{反对称阵\} range σ={反对称阵}
例3: σ : P [ R ] → P [ R ] , f ∈ P [ R ] , σ ( f ) = f ′ \sigma : P[R]\rightarrow P[R],f \in P[R],\sigma(f)=f' σ:P[R]→P[R],f∈P[R],σ(f)=f′: r a n g e σ = P [ R ] range\ \sigma=P[R] range σ=P[R]
例4: σ : P [ R ] → P [ R ] , f ∈ P [ R ] , σ ( f ) = x 3 f \sigma : P[R]\rightarrow P[R],f \in P[R],\sigma(f)=x^3f σ:P[R]→P[R],f∈P[R],σ(f)=x3f: r a n g e σ = P [ R ] range\ \sigma=P[R] range σ=P[R]
2.3.3 维数定理及其推断
维数公式: d i m ( n u l l σ ) + d i m ( r a n g e σ ) = d i m ( U ) dim(null\ \sigma)+dim(range\ \sigma)=dim(U) dim(null σ)+dim(range σ)=dim(U)
⋆ ⋆ ⋆ \star\star\star ⋆⋆⋆证明:
设 α 1 , α 2 , . . . , α k \alpha_1, \alpha_2,...,\alpha_k α1,α2,...,αk是 n u l l σ null\ \sigma null σ的一组基, α 1 , α 2 , . . . , α k , α k + 1 , . . . , α n \alpha_1,\alpha_2,...,\alpha_k,\alpha_{k+1},...,\alpha_n α1,α2,...,αk,αk+1,...,αn是 n u l l σ null\ \sigma null σ扩充到 U U U的一组基。则 r a n g e σ = S p a n ( σ ( α 1 ) , σ ( α 2 ) , . . . , σ ( α k ) , σ ( α k + 1 ) , . . . , σ ( α n ) ) range\ \sigma=Span(\sigma(\alpha_1),\sigma(\alpha_2),...,\sigma(\alpha_k),\sigma(\alpha_{k+1}),...,\sigma(\alpha_n)) range σ=Span(σ(α1),σ(α2),...,σ(αk),σ(αk+1),...,σ(αn))
又 σ ( α 1 ) = σ ( α 2 ) = . . . = σ ( α k ) = 0 \sigma(\alpha_1)=\sigma(\alpha_2)=...=\sigma(\alpha_k)=0 σ(α1)=σ(α2)=...=σ(αk)=0
⇒ r a n g e σ = S p a n ( σ ( α k + 1 ) , . . . , σ ( α n ) ) \Rightarrow range\ \sigma=Span(\sigma(\alpha_{k+1}),...,\sigma(\alpha_n)) ⇒range σ=Span(σ(αk+1),...,σ(αn))
已知: d i m ( n u l l σ ) = k , d i m ( U ) = n dim(null\ \sigma)=k,dim(U)=n dim(null σ)=k,dim(U)=n
则需证明 d i m ( r a n g e σ ) = n − k dim(range\ \sigma)=n-k dim(range σ)=n−k,即证 σ ( α k + 1 ) , . . . , σ ( α n ) \sigma(\alpha_{k+1}),...,\sigma(\alpha_n) σ(αk+1),...,σ(αn)线性无关。
设存在一组系数 x k + 1 , . . . , x n x_{k+1},...,x_n xk+1,...,xn使得 x k + 1 ⋅ σ ( α k + 1 ) + . . . + x n ⋅ σ ( α n ) = 0 x_{k+1}\cdot\sigma(\alpha_{k+1})+...+x_n\cdot\sigma(\alpha_n)=0 xk+1⋅σ(αk+1)+...+xn⋅σ(αn)=0
⟺ σ ( x k + 1 ⋅ α k + 1 ) + . . . + σ ( x n ⋅ α n ) = 0 ⟺ σ ( x k + 1 ⋅ α k + 1 + . . . + x n ⋅ α n ) = 0 ⟺ x k + 1 ⋅ α k + 1 + . . . + x n ⋅ α n ∈ n u l l σ \iff \sigma(x_{k+1}\cdot\alpha_{k+1})+...+\sigma(x_n\cdot\alpha_n)=0\iff \sigma(x_{k+1}\cdot\alpha_{k+1}+...+x_n\cdot\alpha_n)=0\iff x_{k+1}\cdot\alpha_{k+1}+...+x_n\cdot\alpha_n\in null\ \sigma ⟺σ(xk+1⋅αk+1)+...+σ(xn⋅αn)=0⟺σ(xk+1⋅αk+1+...+xn⋅αn)=0⟺xk+1⋅αk+1+...+xn⋅αn∈null σ
⟺ x k + 1 ⋅ α k + 1 + . . . + x n ⋅ α n \iff x_{k+1}\cdot\alpha_{k+1}+...+x_n\cdot\alpha_n ⟺xk+1⋅αk+1+...+xn⋅αn可以用 n u l l σ null\ \sigma null σ中的基线性表示,设 x 1 , x 2 , . . . , x k x_1,x_2,...,x_k x1,x2,...,xk是 x k + 1 ⋅ α k + 1 + . . . + x n ⋅ α n x_{k+1}\cdot\alpha_{k+1}+...+x_n\cdot\alpha_n xk+1⋅αk+1+...+xn⋅αn在基 α 1 , α 2 , . . . , α k \alpha_1, \alpha_2,...,\alpha_k α1,α2,...,αk下的坐标
⇒ x k + 1 ⋅ α k + 1 + . . . + x n ⋅ α n = x 1 ⋅ α 1 + x 2 ⋅ α 2 + . . . + x k ⋅ α k \Rightarrow x_{k+1}\cdot\alpha_{k+1}+...+x_n\cdot\alpha_n=x_1\cdot\alpha_1+x_2\cdot\alpha_2+...+x_k\cdot\alpha_k ⇒xk+1⋅αk+1+...+xn⋅αn=x1⋅α1+x2⋅α2+...+xk⋅αk
⟺ x 1 ⋅ α 1 + x 2 ⋅ α 2 + . . . + x k ⋅ α k − x k + 1 ⋅ α k + 1 − . . . − x n ⋅ α n = 0 \iff x_1\cdot\alpha_1+x_2\cdot\alpha_2+...+x_k\cdot\alpha_k-x_{k+1}\cdot\alpha_{k+1}-...-x_n\cdot\alpha_n=0 ⟺x1⋅α1+x2⋅α2+...+xk⋅αk−xk+1⋅αk+1−...−xn⋅αn=0
又 α 1 , α 2 , . . . , α k , α k + 1 , . . . , α n \alpha_1,\alpha_2,...,\alpha_k,\alpha_{k+1},...,\alpha_n α1,α2,...,αk,αk+1,...,αn是 U U U的一组基 ⇒ α 1 , α 2 , . . . , α k , α k + 1 , . . . , α n \Rightarrow \alpha_1,\alpha_2,...,\alpha_k,\alpha_{k+1},...,\alpha_n ⇒α1,α2,...,αk,αk+1,...,αn线性无关
⇒ x 1 = x 2 = . . . = x k = x k + 1 = . . . = x n = 0 ⇒ σ ( α k + 1 ) , . . . , σ ( α n ) \Rightarrow x_1=x_2=...=x_k=x_{k+1}=...=x_n=0\Rightarrow \sigma(\alpha_{k+1}),...,\sigma(\alpha_n) ⇒x1=x2=...=xk=xk+1=...=xn=0⇒σ(αk+1),...,σ(αn) 线性无关。
证毕。
维数公式的推断
(1) d i m ( U ) > d i m ( V ) ⇒ 一 定 不 是 单 射 dim(U)>dim(V)\Rightarrow 一定不是单射 dim(U)>dim(V)⇒一定不是单射
证明:
已知 d i m ( n u l l σ ) + d i m ( r a n g e σ ) = d i m ( U ) dim(null\ \sigma)+dim(range\ \sigma)=dim(U) dim(null σ)+dim(range σ)=dim(U)
由 d i m ( U ) > d i m ( V ) dim(U)>dim(V) dim(U)>dim(V)
⇒ 0 < d i m ( U ) − d i m ( V ) ≤ d i m ( U ) − d i m ( r a n g e σ ) = d i m ( n u l l σ ) \Rightarrow 0<dim(U)-dim(V) \leq dim(U)-dim(range\ \sigma)=dim(null\ \sigma) ⇒0<dim(U)−dim(V)≤dim(U)−dim(range σ)=dim(null σ)
⇒ U → V 一 定 不 是 单 射 \Rightarrow U\rightarrow V一定不是单射 ⇒U→V一定不是单射
(2) d i m ( U ) < d i m ( V ) ⇒ 一 定 不 是 满 射 dim(U)<dim(V)\Rightarrow 一定不是满射 dim(U)<dim(V)⇒一定不是满射
证明:
已知 d i m ( n u l l σ ) + d i m ( r a n g e σ ) = d i m ( U ) dim(null\ \sigma)+dim(range\ \sigma)=dim(U) dim(null σ)+dim(range σ)=dim(U)
⇒ d i m ( r a n g e σ ) = d i m ( U ) − d i m ( n u l l σ ) ≤ d i m ( U ) < d i m ( V ) \Rightarrow dim(range\ \sigma)=dim(U)-dim(null\ \sigma)\leq dim(U)<dim(V) ⇒dim(range σ)=dim(U)−dim(null σ)≤dim(U)<dim(V)
⇒ U → V 一 定 不 是 满 射 \Rightarrow U\rightarrow V一定不是满射 ⇒U→V一定不是满射
例1:证明对于矩阵 A m × n A_{m\times n} Am×n,当 n > m n>m n>m时 A x = 0 Ax=0 Ax=0一定有非零解
证明:对于 A x = 0 Ax=0 Ax=0相当于是 σ : R n → R m \sigma:R^n\rightarrow R^m σ:Rn→Rm的线性变换,其中 x ∈ R n x\in R^n x∈Rn,则 A x = 0 Ax=0 Ax=0的解就是零空间 n u l l σ null\ \sigma null σ
又 n > m n>m n>m从高维映射到低维一定不是单射 ⇒ \Rightarrow ⇒ A x = 0 Ax=0 Ax=0的解这一零空间一定有除0以外的元素 ⟺ A x = 0 \iff Ax=0 ⟺Ax=0一定存在非零解
例2:证明对于矩阵 A m × n A_{m\times n} Am×n,当 n < m n<m n<m时,存在 b b b使得 A x = b Ax=b Ax=b无解
证明:对于 A x = b Ax=b Ax=b相当于是 σ : R n → R m \sigma:R^n\rightarrow R^m σ:Rn→Rm的线性变换,其中 x ∈ R n , b ∈ R m x\in R^n,b\in R^m x∈Rn,b∈Rm,则使 A x = b Ax=b Ax=b有解的 b b b构成的集合就是像空间 r a n g e σ range\ \sigma range σ
又 n < m n<m n<m从低维映射到高维一定不是满射 ⇒ r a n g e σ < R m ⟺ R m \Rightarrow range\ \sigma<R^m\iff R^m ⇒range σ<Rm⟺Rm中一定存在 b b b不在 使 A x = b Ax=b Ax=b有解的 b b b构成的空间内 ⟺ A x = b \iff Ax=b ⟺Ax=b一定存在 b b b使其无解
例3:已知一个线性变换 σ : R 4 → R 4 \sigma:R^4\rightarrow R^4 σ:R4→R4,找一个 n u l l σ = r a n g e σ null\ \sigma=range\ \sigma null σ=range σ的线性例子
解:记该线性变换为 L ( U ) L(U) L(U),设 e 1 , e 2 , e 3 , e 4 e_1,e_2,e_3,e_4 e1,e2,e3,e4为 U U U中的一组基
令 σ ( e 1 ) = 0 , σ ( e 2 ) = 0 , σ ( e 3 ) = e 1 , σ ( e 4 ) = e 2 \sigma(e_1)=0,\sigma(e_2)=0,\sigma(e_3)=e_1,\sigma(e_4)=e_2 σ(e1)=0,σ(e2)=0,σ(e3)=e1,σ(e4)=e2
证明: r a n g e σ = S p a n ( σ ( e 1 ) , σ ( e 2 ) , σ ( e 3 ) , σ ( e 4 ) ) = S p a n ( 0 , 0 , e 1 , e 2 ) = S p a n ( e 1 , e 2 ) range\ \sigma=Span(\sigma(e_1),\sigma(e_2),\sigma(e_3),\sigma(e_4))=Span(0,0,e_1,e_2)=Span(e_1,e_2) range σ=Span(σ(e1),σ(e2),σ(e3),σ(e4))=Span(0,0,e1,e2)=Span(e1,e2)
又 d i m ( n u l l σ ) = d i m ( U ) − d i m ( r a n g e σ ) = 4 − 2 = 2 dim(null\ \sigma)=dim(U)-dim(range\ \sigma)=4-2=2 dim(null σ)=dim(U)−dim(range σ)=4−2=2
又 σ ( e 1 ) = σ ( e 2 ) = 0 ⇒ e 1 , e 2 ∈ n u l l σ \sigma(e_1)=\sigma(e_2)=0\Rightarrow e_1,e_2\in null\ \sigma σ(e1)=σ(e2)=0⇒e1,e2∈null σ且 e 1 , e 2 e_1,e_2 e1,e2线性无关
⇒ n u l l σ = S p a n ( e 1 , e 2 ) = r a n g e σ \Rightarrow null\ \sigma=Span(e_1,e_2)=range\ \sigma ⇒null σ=Span(e1,e2)=range σ
证毕。
例4:对于线性变换 σ : U → V , d i m ( U ) = 5 , d i m ( V ) = 4 \sigma:U\rightarrow V, dim(U)=5,dim(V)=4 σ:U→V,dim(U)=5,dim(V)=4,是否存线性变换使得 d i m ( n u l l σ ) = d i m ( r a n g e σ ) dim(null\ \sigma)=dim(range\ \sigma) dim(null σ)=dim(range σ)?
解:不存在,因为 U U U为奇数,而 d i m ( n u l l σ ) + d i m ( r a n g e σ ) = d i m ( U ) ⇒ dim(null\ \sigma)+dim(range\ \sigma)=dim(U)\Rightarrow dim(null σ)+dim(range σ)=dim(U)⇒不可能存在 d i m ( n u l l σ ) = d i m ( r a n g e σ ) dim(null\ \sigma)=dim(range\ \sigma) dim(null σ)=dim(range σ)
例5:对于线性变换 σ : U → U \sigma:U\rightarrow U σ:U→U,是否满足 n u l l σ ⊕ r a n g e σ = U null\ \sigma\oplus range\ \sigma=U null σ⊕range σ=U?
解:不一定满足,示例即例3
2.3.4 乘积空间
定义:设 U 1 , U 2 , . . . , U k U_1,U_2,...,U_k U1,U2,...,Uk是 F F F上的线性空间,则 U 1 × U 2 × . . . × U k = { ( α 1 , α 2 , . . . , α k ) , α i ∈ U i } U_1\times U_2\times ...\times U_k=\{(\alpha_1,\alpha_2,...,\alpha_k),\alpha_i\in U_i\} U1×U2×...×Uk={(α1,α2,...,αk),αi∈Ui}
例1: R 2 × P 3 [ R ] = { ( a 1 , a 2 ) , a 3 x 3 + a 4 x 2 + a 5 x + a 6 , a i ∈ R } , d i m ( R 2 × P 3 [ R ] ) = 6 R^2\times P_3[R]=\{(a_1,a_2),a_3x^3+a_4x^2+a_5x+a_6,a_i\in R\},dim(R^2\times P_3[R])=6 R2×P3[R]={(a1,a2),a3x3+a4x2+a5x+a6,ai∈R},dim(R2×P3[R])=6
例2:设 U 1 , U 2 , . . . , U k U_1,U_2,...,U_k U1,U2,...,Uk为 V V V的子空间,写出一个乘积空间 U 1 × U 2 × . . . × U k U_1\times U_2\times...\times U_k U1×U2×...×Uk到和空间 U 1 + U 2 + . . . + U k U_1+ U_2+...+U_k U1+U2+...+Uk的线性变换
如 σ ( α 1 , α 2 , . . . , α k ) = α 1 + α 2 + . . . + α k , α i ∈ U i \sigma(\alpha_1,\alpha_2,...,\alpha_k)=\alpha_1+\alpha_2+...+\alpha_k,\alpha_i\in U_i σ(α1,α2,...,αk)=α1+α2+...+αk,αi∈Ui
2.3.5 商空间
定义:设 U ∈ V , v ∈ V , v + U = { v + u , u ∈ U } U\in V,v\in V,v+U=\{v+u,u\in U\} U∈V,v∈V,v+U={v+u,u∈U},则把 V / U = { v + U , v ∈ V } V/U=\{v+U,v\in V\} V/U={v+U,v∈V}称作商空间。
商空间的加法: v 1 + U , v 2 + U ∈ V / U v_1+U,v_2+U\in V/U v1+U,v2+U∈V/U,定义商空间的加法运算为 ( v 1 + v 2 ) + U (v_1+v_2)+U (v1+v2)+U
商空间的数乘: v + U ∈ V / U , k ∈ F v+U\in V/U,k\in F v+U∈V/U,k∈F,定义商空间的加法运算为 ( k ⋅ v ) + U (k\cdot v)+U (k⋅v)+U
性质
∀ α , β ∈ V / U , α − β ∈ U ⟺ α + U = β + U ⟺ α + U ∩ β + U ≠ ∅ \forall \alpha,\beta\in V/U,\alpha-\beta\in U\iff\alpha+U=\beta+U\iff \alpha+U\cap\beta+U\not=\empty ∀α,β∈V/U,α−β∈U⟺α+U=β+U⟺α+U∩β+U=∅
证明:
(1) α − β ∈ U ⇒ α + U = β + U \alpha-\beta\in U\Rightarrow\alpha+U=\beta+U α−β∈U⇒α+U=β+U
α + U = β + ( α − β ) + U ∈ β + U \alpha+U=\beta+(\alpha-\beta)+U\in\beta+U α+U=β+(α−β)+U∈β+U
β + U = α + ( β − α ) + U ∈ α + U \beta+U=\alpha+(\beta-\alpha)+U\in\alpha+U β+U=α+(β−α)+U∈α+U
⇒ α + U = β + U \Rightarrow \alpha+U=\beta+U ⇒α+U=β+U
(2) α + U = β + U ⇒ α + U ∩ β + U ≠ ∅ \alpha+U=\beta+U\Rightarrow \alpha+U\cap\beta+U\not=\empty α+U=β+U⇒α+U∩β+U=∅
相等的两个非空集合相交一定不为空
(3) α + U ∩ β + U ≠ ∅ ⇒ α − β ∈ U \alpha+U\cap\beta+U\not=\empty\Rightarrow\alpha-\beta\in U α+U∩β+U=∅⇒α−β∈U
存在 u 1 , u 2 ∈ U , α + u 1 = β + u 2 u_1,u_2\in U,\alpha+u_1=\beta+u_2 u1,u2∈U,α+u1=β+u2
⇒ α − β = u 2 − u 1 ∈ U \Rightarrow \alpha-\beta=u_2-u_1\in U ⇒α−β=u2−u1∈U
证毕。
U , V U,V U,V是有限维线性空间, S ∈ L ( V , W ) , T ∈ L ( U , V ) S\in L(V,W),T\in L(U,V) S∈L(V,W),T∈L(U,V)
证 d i m ( n u l l S T ) ≤ d i m ( n u l l S ) + d i m ( n u l l T ) dim(null\ ST) \le dim(null\ S)+dim(null\ T) dim(null ST)≤dim(null S)+dim(null T)
提示:零(核)空间与基下矩阵的零空间同构,零(核)空间的维数和矩阵的零空间维数相同。
理解商空间
- 具体化到 V ∈ R 2 , U ∈ R V\in R^2,U\in R V∈R2,U∈R,也就是直线和平面。
- 等价于把 V V V向 U U U的垂直方向压缩, U U U被压缩为 0 0 0
3 线性变换矩阵及其应用
3.1 线性变换矩阵含义
设
α
1
,
α
2
,
.
.
.
,
α
n
\alpha_1,\alpha_2,...,\alpha_n
α1,α2,...,αn是
U
U
U的基,
β
1
,
β
2
,
.
.
.
,
β
m
\beta_1,\beta_2,...,\beta_m
β1,β2,...,βm是
V
V
V的基
{
σ
(
α
1
)
=
a
11
β
1
+
a
21
β
2
+
.
.
.
+
a
m
1
β
m
σ
(
α
2
)
=
a
12
β
1
+
a
22
β
2
+
.
.
.
+
a
m
2
β
m
.
.
σ
(
α
n
)
=
a
1
n
β
1
+
a
2
n
β
2
+
.
.
.
+
a
m
n
β
m
\begin{cases} \sigma(\alpha_1)=a_{11}\beta_1+a_{21}\beta_2+...+a_{m1}\beta_m\\ \sigma(\alpha_2)=a_{12}\beta_1+a_{22}\beta_2+...+a_{m2}\beta_m\\ \ \ .\\ \ \ .\\ \sigma(\alpha_n)=a_{1n}\beta_1+a_{2n}\beta_2+...+a_{mn}\beta_m \end{cases}
⎩⎪⎪⎪⎪⎪⎪⎨⎪⎪⎪⎪⎪⎪⎧σ(α1)=a11β1+a21β2+...+am1βmσ(α2)=a12β1+a22β2+...+am2βm . .σ(αn)=a1nβ1+a2nβ2+...+amnβm
⇒
σ
(
α
1
,
α
2
,
.
.
.
,
α
n
)
=
(
σ
(
α
1
)
,
σ
(
α
2
)
,
.
.
.
,
σ
(
α
n
)
)
=
(
β
1
,
β
2
,
.
.
.
,
β
m
)
⋅
(
a
11
a
12
.
.
.
a
1
n
a
21
a
22
.
.
.
a
2
n
.
.
.
.
.
.
.
.
.
.
.
.
a
m
1
a
m
2
.
.
.
a
m
n
)
\Rightarrow \sigma(\alpha_1,\alpha_2,...,\alpha_n)=(\sigma(\alpha_1),\sigma(\alpha_2),...,\sigma(\alpha_n))=(\beta_1,\beta_2,...,\beta_m)\cdot \begin{pmatrix}a_{11}&a_{12}&...&a_{1n}\\a_{21}&a_{22}&...&a_{2n}\\...&...&...&...\\a_{m1}&a_{m2}&...&a_{mn}\end{pmatrix}
⇒σ(α1,α2,...,αn)=(σ(α1),σ(α2),...,σ(αn))=(β1,β2,...,βm)⋅⎝⎜⎜⎛a11a21...am1a12a22...am2............a1na2n...amn⎠⎟⎟⎞
令 A = ( a 11 a 12 . . . a 1 n a 21 a 22 . . . a 2 n . . . . . . . . . . . . a m 1 a m 2 . . . a m n ) A=\begin{pmatrix}a_{11}&a_{12}&...&a_{1n}\\a_{21}&a_{22}&...&a_{2n}\\...&...&...&...\\a_{m1}&a_{m2}&...&a_{mn}\end{pmatrix} A=⎝⎜⎜⎛a11a21...am1a12a22...am2............a1na2n...amn⎠⎟⎟⎞成为线性变换 σ \sigma σ在给定两组基下的变换矩阵。
⋆ ⋆ ⋆ \star\star\star ⋆⋆⋆例1:已知线性变换 σ : R 3 → R 2 , σ ( ( x , y , z ) T ) = ( 2 x + y + z , x − z ) T \sigma:R^3\rightarrow R^2,\sigma((x,y,z)^T)=(2x+y+z,x-z)^T σ:R3→R2,σ((x,y,z)T)=(2x+y+z,x−z)T,求 U , V U,V U,V空间对应标准基的变换矩阵
解:设 e 1 , e 2 , e 3 ∈ R 3 e_1,e_2,e_3\in R^3 e1,e2,e3∈R3是 U U U的标准基, E 1 , E 2 ∈ R 2 E_1,E_2\in R^2 E1,E2∈R2是 V V V的标准基
⇒ σ ( e 1 ) = ( 2 , 1 ) T , σ ( e 2 ) = ( 1 , 0 ) T , σ ( e 3 ) = ( 1 , − 1 ) T \Rightarrow \sigma(e_1)=(2,1)^T,\sigma(e_2)=(1,0)^T,\sigma(e_3)=(1,-1)^T ⇒σ(e1)=(2,1)T,σ(e2)=(1,0)T,σ(e3)=(1,−1)T
⇒ ( σ ( e 1 ) , σ ( e 2 ) , σ ( e 3 ) ) = ( E 1 , E 2 ) ⋅ A \Rightarrow (\sigma(e_1),\sigma(e_2),\sigma(e_3))=(E_1,E_2)\cdot A ⇒(σ(e1),σ(e2),σ(e3))=(E1,E2)⋅A
⇒ A = ( σ ( e 1 ) , σ ( e 2 ) , σ ( e 3 ) ) = [ 2 1 1 1 0 − 1 ] \Rightarrow A=(\sigma(e_1),\sigma(e_2),\sigma(e_3))=\begin{bmatrix}2&1&1\\1&0&-1\end{bmatrix} ⇒A=(σ(e1),σ(e2),σ(e3))=[21101−1]
作业一:已知线性变换 σ : U → U , α 1 , α 2 , α 3 \sigma:U\rightarrow U,\alpha_1,\alpha_2,\alpha_3 σ:U→U,α1,α2,α3是 U U U的一组基,且对应变换矩阵满足 σ ( α 1 , α 2 , α 3 ) = ( α 1 , α 2 , α 3 ) ⋅ [ a 11 a 12 a 13 a 21 a 22 a 23 a 31 a 32 a 33 ] \sigma(\alpha_1,\alpha_2,\alpha_3)=(\alpha_1,\alpha_2,\alpha_3)\cdot\begin{bmatrix}a_{11}&a_{12}&a_{13}\\a_{21}&a_{22}&a_{23}\\a_{31}&a_{32}&a_{33}\end{bmatrix} σ(α1,α2,α3)=(α1,α2,α3)⋅⎣⎡a11a21a31a12a22a32a13a23a33⎦⎤,求在另一组基 ( 3 α 3 , α 2 , 4 α 1 ) (3\alpha_3,\alpha_2,4\alpha_1) (3α3,α2,4α1)下,该线性变换对应的变换矩阵。
3.2 线性变换矩阵性质
⋆ ⋆ ⋆ \star\star\star ⋆⋆⋆性质1:同一线性变换在不同基下的矩阵相似
性质2:加法、数乘、嵌套变换下的矩阵的关系
σ , τ ∈ L ( U , V ) , σ ≠ τ , ( α 1 , α 2 , . . . , α n ) 与 ( β 1 , β 2 , . . . , β m ) 分 别 是 U , V 的 一 组 基 , k ∈ F \sigma,\tau\in L(U,V),\sigma\not=\tau,(\alpha_1,\alpha_2,...,\alpha_n)与(\beta_1, \beta_2,...,\beta_m)分别是U,V的一组基,k\in F σ,τ∈L(U,V),σ=τ,(α1,α2,...,αn)与(β1,β2,...,βm)分别是U,V的一组基,k∈F
σ ( α 1 , α 2 , . . . , α n ) = ( β 1 , β 2 , . . . , β m ) ⋅ A \sigma(\alpha_1,\alpha_2,...,\alpha_n)=(\beta_1,\beta_2,...,\beta_m)\cdot A σ(α1,α2,...,αn)=(β1,β2,...,βm)⋅A
τ ( α 1 , α 2 , . . . , α n ) = ( β 1 , β 2 , . . . , β m ) ⋅ B \tau(\alpha_1,\alpha_2,...,\alpha_n)=(\beta_1,\beta_2,...,\beta_m)\cdot B τ(α1,α2,...,αn)=(β1,β2,...,βm)⋅B
- ( σ + τ ) ( α 1 , α 2 , . . . , α n ) = ( β 1 , β 2 , . . . , β m ) ⋅ ( A + B ) (\sigma+\tau)(\alpha_1,\alpha_2,...,\alpha_n)=(\beta_1,\beta_2,...,\beta_m)\cdot(A+B) (σ+τ)(α1,α2,...,αn)=(β1,β2,...,βm)⋅(A+B)
- ( k ⋅ σ ) ( α 1 , α 2 , . . . , α n ) = ( β 1 , β 2 , . . . , β m ) ⋅ ( k ⋅ A ) (k\cdot\sigma)(\alpha_1,\alpha_2,...,\alpha_n)=(\beta_1,\beta_2,...,\beta_m)\cdot(k\cdot A) (k⋅σ)(α1,α2,...,αn)=(β1,β2,...,βm)⋅(k⋅A)
σ 1 ∈ L ( U , V ) , σ 2 ∈ L ( V , W ) , ( α 1 , α 2 , . . . , α n ) , ( β 1 , β 2 , . . . , β m ) , ( γ 1 , γ 2 , . . . , γ p ) 分 别 为 U , V , W 的 一 组 基 \sigma_1\in L(U,V),\sigma_2\in L(V,W),(\alpha_1,\alpha_2,...,\alpha_n),(\beta_1,\beta_2,...,\beta_m),(\gamma_1,\gamma_2,...,\gamma_p)分别为U,V,W的一组基 σ1∈L(U,V),σ2∈L(V,W),(α1,α2,...,αn),(β1,β2,...,βm),(γ1,γ2,...,γp)分别为U,V,W的一组基
σ 1 ( α 1 , α 2 , . . . , α n ) = ( β 1 , β 2 , . . . , β m ) ⋅ A m × n \sigma_1(\alpha_1,\alpha_2,...,\alpha_n)=(\beta_1,\beta_2,...,\beta_m)\cdot A_{m\times n} σ1(α1,α2,...,αn)=(β1,β2,...,βm)⋅Am×n
σ 2 ( β 1 , β 2 , . . . , β m ) = ( γ 1 , γ 2 , . . . , γ p ) ⋅ B p × m \sigma_2(\beta_1,\beta_2,...,\beta_m)=(\gamma_1,\gamma_2,...,\gamma_p)\cdot B_{p\times m} σ2(β1,β2,...,βm)=(γ1,γ2,...,γp)⋅Bp×m
- ( σ 2 ⋅ σ 1 ) ( α 1 , α 2 , . . . , α n ) = ( γ 1 , γ 2 , . . . , γ p ) ⋅ ( B p × m ⋅ A m × n ) (\sigma_2\cdot\sigma_1)(\alpha_1,\alpha_2,...,\alpha_n)=(\gamma_1,\gamma_2,...,\gamma_p)\cdot(B_{p\times m}\cdot A_{m\times n}) (σ2⋅σ1)(α1,α2,...,αn)=(γ1,γ2,...,γp)⋅(Bp×m⋅Am×n)
3.3 线性变换矩阵应用
用于求零空间与像空间
对于一个线性变换 σ : U → V , d i m ( U ) = n , d i m ( V ) = m \sigma:U\rightarrow V,dim(U)=n,dim(V)=m σ:U→V,dim(U)=n,dim(V)=m
α 1 , α 2 , . . . , α n \alpha_1,\alpha_2,...,\alpha_n α1,α2,...,αn与 β 1 , β 2 , . . . , β m \beta_1,\beta_2,...,\beta_m β1,β2,...,βm分别是 U , V U,V U,V的一组基
A m × n A_{m\times n} Am×n是两组基下的线性变换矩阵 ⇒ σ ( α 1 , α 2 , . . . , α n ) = ( β 1 , β 2 , . . . , β m ) ⋅ A \Rightarrow \sigma(\alpha_1,\alpha_2,...,\alpha_n)=(\beta_1,\beta_2,...,\beta_m)\cdot A ⇒σ(α1,α2,...,αn)=(β1,β2,...,βm)⋅A
对于任意的 α ∈ U \alpha\in U α∈U,存在坐标 X X X使得 α = ( α 1 , α 2 , . . . , α n ) ⋅ X \alpha=(\alpha_1,\alpha_2,...,\alpha_n)\cdot X α=(α1,α2,...,αn)⋅X
⇒ σ ( α ) = σ ( ( α 1 , α 2 , . . . , α n ) ⋅ X ) = σ ( α 1 , α 2 , . . . , α n ) ⋅ X = ( β 1 , β 2 , . . . , β m ) ⋅ A ⋅ X \Rightarrow \sigma(\alpha)=\sigma((\alpha_1,\alpha_2,...,\alpha_n)\cdot X)=\sigma(\alpha_1,\alpha_2,...,\alpha_n)\cdot X=(\beta_1,\beta_2,...,\beta_m)\cdot A\cdot X ⇒σ(α)=σ((α1,α2,...,αn)⋅X)=σ(α1,α2,...,αn)⋅X=(β1,β2,...,βm)⋅A⋅X
-
零空间
σ ( α ) = 0 ⟺ A X = 0 \sigma(\alpha)=0\iff AX=0 σ(α)=0⟺AX=0
⇒ \Rightarrow ⇒求解 N ( A ) N(A) N(A),解得的 X X X回乘 ( α 1 , α 2 , . . . , α n ) (\alpha_1,\alpha_2,...,\alpha_n) (α1,α2,...,αn)得到零空间
-
像空间
A X AX AX表示 A A A的列空间 C ( A ) C(A) C(A),即像空间 r a n g e σ = ( β 1 , β 2 , . . . , β m ) ⋅ C ( A ) range\ \sigma=(\beta_1,\beta_2,...,\beta_m)\cdot C(A) range σ=(β1,β2,...,βm)⋅C(A)
4 线性变换的逆变换
4.1 逆变换定义与性质
定义:若 σ ∈ L ( U , V ) , τ ∈ L ( V , U ) \sigma\in L(U,V),\tau\in L(V,U) σ∈L(U,V),τ∈L(V,U)且 σ ⋅ τ = τ ⋅ σ = I , I \sigma\cdot\tau=\tau\cdot\sigma=I,I σ⋅τ=τ⋅σ=I,I表示恒等变换,则称 τ \tau τ是 σ \sigma σ的逆变换,或者说 σ \sigma σ与 τ \tau τ同构
(有时也称U与V空间同构,两空间同构的含义是存在至少一组可逆变换)
性质
- 如果 σ ∈ L ( U , V ) \sigma\in L(U,V) σ∈L(U,V)存在逆变换 τ \tau τ则该逆变换唯一。
证明:
设 σ ∈ L ( U , V ) \sigma\in L(U,V) σ∈L(U,V)存在两个逆矩阵 τ 1 , τ 2 \tau_1,\tau_2 τ1,τ2
则 τ 1 = τ 1 ⋅ I V → V = τ 1 ⋅ ( σ ⋅ τ 2 ) = ( τ 1 ⋅ σ ) ⋅ τ 2 = I U → U ⋅ τ 2 = τ 2 \tau_1=\tau_1\cdot I_{V\rightarrow V}=\tau_1\cdot(\sigma\cdot\tau_2)=(\tau_1\cdot\sigma)\cdot\tau_2=I_{U\rightarrow U}\cdot\tau_2=\tau_2 τ1=τ1⋅IV→V=τ1⋅(σ⋅τ2)=(τ1⋅σ)⋅τ2=IU→U⋅τ2=τ2
即 τ 1 = τ 2 \tau_1=\tau_2 τ1=τ2,逆变换唯一
证毕。
- 一个线性变换存在可逆变换(同构) ⟺ \iff ⟺该变换既是单射又是满射。
证明:
设 σ ∈ L ( U , V ) \sigma\in L(U,V) σ∈L(U,V),若存在逆变换,则记作 σ − 1 \sigma^{-1} σ−1
(1)一个线性变换存在逆变换 ⇒ \Rightarrow ⇒单射
设 α , β ∈ U , σ ( α ) = σ ( β ) \alpha,\beta\in U,\sigma(\alpha)=\sigma(\beta) α,β∈U,σ(α)=σ(β)
⇒ σ − 1 ( σ ( α ) ) = σ − 1 ( σ ( β ) ) ⇒ α = β \Rightarrow \sigma^{-1}(\sigma(\alpha))=\sigma^{-1}(\sigma(\beta))\Rightarrow\alpha=\beta ⇒σ−1(σ(α))=σ−1(σ(β))⇒α=β
⇒ 该 命 题 成 立 \Rightarrow 该命题成立 ⇒该命题成立
(2)一个线性变换存在逆变换 ⇒ \Rightarrow ⇒满射
∀ β ∈ V , ∃ α = σ − 1 ( β ) ∈ U , 使 得 σ ( α ) = σ ( σ − 1 ( β ) ) = β \forall \beta\in V,\exist\alpha=\sigma^{-1}(\beta)\in U,使得\sigma(\alpha)=\sigma(\sigma^{-1}(\beta))=\beta ∀β∈V,∃α=σ−1(β)∈U,使得σ(α)=σ(σ−1(β))=β
⇒ \Rightarrow ⇒ 该命题成立
(3)一个线性变换既是单射又是满射 ⇒ \Rightarrow ⇒该变换存在逆变换
设 σ ∈ L ( U , V ) \sigma\in L(U,V) σ∈L(U,V)既是单射又是满射
则存在 ∀ α ∈ U , ∃ 唯 一 β , 使 得 σ ( α ) = β \forall \alpha\in U,\exist 唯一\beta,使得\sigma(\alpha)=\beta ∀α∈U,∃唯一β,使得σ(α)=β
反之存在 ∀ β ∈ V , ∃ 唯 一 α , 使 得 τ ( β ) = α , τ \forall \beta\in V,\exist 唯一\alpha,使得\tau(\beta)=\alpha,\tau ∀β∈V,∃唯一α,使得τ(β)=α,τ表示 V → U V\rightarrow U V→U的一个映射
⇒ σ ⋅ τ ( β ) = σ ( α ) = β \Rightarrow \sigma\cdot\tau(\beta)=\sigma(\alpha)=\beta ⇒σ⋅τ(β)=σ(α)=β
τ ⋅ σ ( α ) = τ ( β ) = α \tau\cdot\sigma(\alpha)=\tau(\beta)=\alpha τ⋅σ(α)=τ(β)=α
⇒ τ ⋅ σ = σ ⋅ τ = I \Rightarrow \tau\cdot\sigma=\sigma\cdot\tau=I ⇒τ⋅σ=σ⋅τ=I
又设 β 1 = σ ( α 1 ) , β 2 = σ ( α 2 ) ∈ V , k ∈ F \beta_1=\sigma(\alpha_1),\beta_2=\sigma(\alpha_2)\in V,k\in F β1=σ(α1),β2=σ(α2)∈V,k∈F
⇒ τ ( β 1 + β 2 ) = τ ( σ ( α 1 ) + σ ( α 2 ) ) = τ ( σ ( α 1 + α 2 ) ) = ( τ ⋅ σ ) ( α 1 + α 2 ) = α 1 + α 2 = τ ( β 1 ) + τ ( β 2 ) \Rightarrow \tau(\beta_1+\beta_2)=\tau(\sigma(\alpha_1)+\sigma(\alpha_2))=\tau(\sigma(\alpha_1+\alpha_2))=(\tau\cdot\sigma)(\alpha_1+\alpha_2)=\alpha_1+\alpha_2=\tau(\beta_1)+\tau(\beta_2) ⇒τ(β1+β2)=τ(σ(α1)+σ(α2))=τ(σ(α1+α2))=(τ⋅σ)(α1+α2)=α1+α2=τ(β1)+τ(β2)
τ ( k ⋅ β 1 ) = τ ( k ⋅ σ ( α 1 ) ) = τ ( σ ( k ⋅ α 1 ) ) = ( τ ⋅ σ ) ( k ⋅ α 1 ) = k ⋅ α 1 = k ⋅ τ ( β 1 ) \tau(k\cdot\beta_1)=\tau(k\cdot\sigma(\alpha_1))=\tau(\sigma(k\cdot\alpha_1))=(\tau\cdot\sigma)(k\cdot\alpha_1)=k\cdot\alpha_1=k\cdot\tau(\beta_1) τ(k⋅β1)=τ(k⋅σ(α1))=τ(σ(k⋅α1))=(τ⋅σ)(k⋅α1)=k⋅α1=k⋅τ(β1)
⇒ τ \Rightarrow \tau ⇒τ是线性变换
综上, τ \tau τ满足线性变换的逆变换的定义 ⇒ τ \Rightarrow \tau ⇒τ是 σ \sigma σ的逆变换
⇒ \Rightarrow ⇒ 该命题成立
- U , V U,V U,V两个线性空间同构 ⟺ d i m ( U ) = d i m ( V ) \iff dim(U)=dim(V) ⟺dim(U)=dim(V)
证明:
(1) U , V U,V U,V两个线性空间存在可逆变换 ⇒ d i m ( U ) = d i m ( V ) \Rightarrow dim(U)=dim(V) ⇒dim(U)=dim(V)
设 σ ∈ L ( U , V ) \sigma\in L(U,V) σ∈L(U,V)是可逆变换,逆变换用 σ − 1 \sigma^{-1} σ−1表示 ⇒ σ \Rightarrow\sigma ⇒σ既是单射又是满射
d i m ( n u l l σ ) = 0 , d i m ( r a n g e σ ) = d i m ( V ) dim(null\ \sigma)=0,dim(range\ \sigma)=dim(V) dim(null σ)=0,dim(range σ)=dim(V)
由维数定理得 dim ( U ) = d i m ( n u l l σ ) + d i m ( r a n g e σ ) = d i m ( V ) \dim(U)=dim(null\ \sigma)+dim(range\ \sigma)=dim(V) dim(U)=dim(null σ)+dim(range σ)=dim(V)
(2) d i m ( U ) = d i m ( V ) ⇒ U , V dim(U)=dim(V)\Rightarrow U,V dim(U)=dim(V)⇒U,V两个线性空间存在可逆变换
设 ( α 1 , α 2 , . . . , α n ) , ( β 1 , β 2 , . . . , β n ) (\alpha_1,\alpha_2,...,\alpha_n),(\beta_1,\beta_2,...,\beta_n) (α1,α2,...,αn),(β1,β2,...,βn)是 U , V U,V U,V的一组基,其中 σ ( α i ) = β i , σ ∈ L ( U , V ) \sigma(\alpha_i)=\beta_i,\sigma\in L(U,V) σ(αi)=βi,σ∈L(U,V)
r a n g e σ = S p a n ( σ ( α 1 ) , σ ( α 2 ) , . . . , σ ( α n ) = S p a n ( β 1 , β 2 , . . . , β n ) = V range\ \sigma=Span(\sigma(\alpha_1),\sigma(\alpha_2),...,\sigma(\alpha_n)=Span(\beta_1,\beta_2,...,\beta_n)=V range σ=Span(σ(α1),σ(α2),...,σ(αn)=Span(β1,β2,...,βn)=V
⇒ d i m ( r a n g e σ ) = d i m ( V ) ⇒ σ \Rightarrow dim(range\ \sigma)=dim(V)\Rightarrow\sigma ⇒dim(range σ)=dim(V)⇒σ是满射
⇒ d i m ( n u l l σ ) = d i m ( U ) − d i m ( r a n g e σ ) = d i m ( U ) − d i m ( V ) = 0 ⇒ σ \Rightarrow dim(null\ \sigma)=dim(U)-dim(range\ \sigma)=dim(U)-dim(V)=0\Rightarrow\sigma ⇒dim(null σ)=dim(U)−dim(range σ)=dim(U)−dim(V)=0⇒σ是单射
- 若 σ ∈ L ( U ) \sigma\in L(U) σ∈L(U)是有限维空间且是单射/满射 ⇒ \Rightarrow ⇒满射/单射
证明:
(1)若 σ ∈ L ( U ) \sigma\in L(U) σ∈L(U)是有限维空间且是单射 ⇒ \Rightarrow ⇒满射
单射 ⇒ d i m ( n u l l σ ) = 0 \Rightarrow dim(null\ \sigma)=0 ⇒dim(null σ)=0
d i m ( r a n g e σ ) = d i m ( U ) − d i m ( n u l l σ ) = d i m ( U ) ⇒ dim(range\ \sigma)=dim(U)-dim(null\ \sigma)=dim(U)\Rightarrow dim(range σ)=dim(U)−dim(null σ)=dim(U)⇒满射
(2)若 σ ∈ L ( U ) \sigma\in L(U) σ∈L(U)是有限维空间且是满射 ⇒ \Rightarrow ⇒单射
满射 ⇒ d i m ( r a n g e σ ) = d i m ( U ) \Rightarrow dim(range\ \sigma)=dim(U) ⇒dim(range σ)=dim(U)
d i m ( n u l l σ ) = d i m ( U ) − d i m ( r a n g e σ ) = 0 ⇒ dim(null\ \sigma)=dim(U)-dim(range\ \sigma)=0\Rightarrow dim(null σ)=dim(U)−dim(range σ)=0⇒单射
证毕。
4.2 逆变换应用
V = F n V=F^n V=Fn与 F F F数域的 n n n维线性空间 U U U同构 ⇒ \Rightarrow ⇒ σ : U → V \sigma:U\rightarrow V σ:U→V可以用 U U U的一组基的坐标 → V \rightarrow V →V来映射
X = F m × n / F n × n X=F^{m\times n}/F^{n\times n} X=Fm×n/Fn×n与 F F F数域中的 W = L ( U , V ) / L ( U ) W=L(U,V)/L(U) W=L(U,V)/L(U)同构 ⇒ σ : W → X \Rightarrow \sigma:W\rightarrow X ⇒σ:W→X可以用 W W W在两组基下的矩阵 → X \rightarrow X →X来映射
零空间 n u l l σ null\ \sigma null σ和基下矩阵的零空间 N ( A ) N(A) N(A)同构;
像空间 r a n g e σ range\ \sigma range σ和基下矩阵的列空间 C ( A ) C(A) C(A)同构;