目录
- 约定
- ω = e − 2 π N \omega = e^{- \frac{2\pi}{N}} ω=e−N2π
- ω i = e − 2 i π N \omega_i = e^{- \frac{2i\pi}{N}} ωi=e−N2iπ
代数基础
- 矩阵与向量
- 零空间 N ( A ) N(A) N(A)
- 像空间 R ( A ) R(A) R(A)
- (列)秩
- r a n k ( A ) = r a n k ( A T ) = r a n k ( A T A ) = r a n k ( A A T ) {\mathrm{rank}}(A) = {\mathrm{rank}}(A^T) = {\mathrm{rank}}(A^TA) = {\mathrm{rank}}(AA^T) rank(A)=rank(AT)=rank(ATA)=rank(AAT)
- r a n k ( A ) = r a n k ( A H ) = r a n k ( A H A ) = r a n k ( A A H ) {\mathrm{rank}}(A) = {\mathrm{rank}}(A^H) = {\mathrm{rank}}(A^HA) = {\mathrm{rank}}(AA^H) rank(A)=rank(AH)=rank(AHA)=rank(AAH)
- 线性空间
- 线性无关
- 基底张成原空间
- 基底表示的唯一性
- 线性空间的运算
- S ∩ T S \cap T S∩T
- S + T S + T S+T
- S ⊕ T S \oplus T S⊕T
- 范数空间&内积空间
- 正交性
- 贝塞尔不等式
- 帕塞瓦尔等式
- 内积空间 ⟹ \implies ⟹ 范数空间
- ∥ x ∥ = ⟨ x , x ⟩ \|x\| = \sqrt{\langle{x},{x}\rangle} ∥x∥=⟨x,x⟩
- 范数空间 ⟹ \implies ⟹ 内积空间
- Reproducing Kernel Hilbert Space
- Mercer Theorem / Moore Aronszajn Theorem
- Jordan von Neumann Theorem
- 正交性
内积空间
- ⟨ x , k y ⟩ ⩽ 1 2 ( ⟨ x , x ⟩ + k 2 ⟨ y , y ⟩ ) \langle{x},{ky}\rangle \leqslant \frac{1}{2} \left( \langle{x},{x}\rangle + k^2 \langle{y},{y}\rangle \right) ⟨x,ky⟩⩽21(⟨x,x⟩+k2⟨y,y⟩)
- ⟨ x , y ⟩ ⩽ ⟨ x , x ⟩ + ⟨ y , y ⟩ 2 \langle{x},{y}\rangle \leqslant \frac{\langle{x},{x}\rangle + \langle{y},{y}\rangle}{2} ⟨x,y⟩⩽2⟨x,x⟩+⟨y,y⟩
- ⟨ x , y ⟩ ⩽ ⟨ x , x ⟩ ⟨ y , y ⟩ \langle{x},{y}\rangle \leqslant \sqrt{\langle{x},{x}\rangle \langle{y},{y}\rangle} ⟨x,y⟩⩽⟨x,x⟩⟨y,y⟩
算子范数
- ∥ A B ∥ ⩽ ∥ A ∥ ∥ B ∥ \|AB\| \leqslant \|A\| \|B\| ∥AB∥⩽∥A∥∥B∥
- ∥ A ∥ 1 = max j ∑ i ∣ A i j ∣ \|A\|_1 = \max\limits_j \sum\limits_i |A_{ij}| ∥A∥1=jmaxi∑∣Aij∣
- ∥ A ∥ ∞ = max i ∑ j ∣ A i j ∣ \|A\|_\infty = \max\limits_i \sum\limits_j |A_{ij}| ∥A∥∞=imaxj∑∣Aij∣
- ∥ A ∥ 2 = max k σ k ( A ) \|A\|_2 = \max\limits_{k} \sigma_k(A) ∥A∥2=kmaxσk(A)
- c o n d ( A ) = ∥ A ∥ ∥ A − 1 ∥ {\mathrm{cond}}(A) = \|A\| \|A^{-1}\| cond(A)=∥A∥∥A−1∥
- c o n d 2 ( A ) = σ max ( A ) σ min ( A ) = σ 1 ( A ) σ n ( A ) {\mathrm{cond}}_2(A) = \frac{\sigma_{\max}(A)}{\sigma_{\min}(A)} = \frac{\sigma_1(A)}{\sigma_n(A)} cond2(A)=σmin(A)σmax(A)=σn(A)σ1(A)
- ∥ A ∥ F = ∑ k σ k 2 ( A ) \|A\|_F = \sqrt{\sum\limits_{k} \sigma_k^2(A)} ∥A∥F=k∑σk2(A)
矩阵基础
- (共轭)转置与乘积
- ( A B ) T = B T A T (AB)^T = B^TA^T (AB)T=BTAT
- ( A B ) H = B H A H (AB)^H = B^HA^H (AB)H=BHAH
- 二次型&(半)正定性
- x H A x ⩾ 0 ⟺ A ≽ 0 x^HAx \geqslant 0 \iff A \succcurlyeq 0 xHAx⩾0⟺A≽0
- A ≽ 0 ⟺ λ i ( A ) ⩾ 0 A \succcurlyeq 0 \iff \lambda_i(A) \geqslant 0 A≽0⟺λi(A)⩾0
- 矩阵正定当且仅当顺序主子式均大于零(Sylvester’s criterion)
- 求逆
- ( A B ) − 1 = B − 1 A − 1 (AB)^{-1} = B^{-1}A^{-1} (AB)−1=B−1A−1
- 行列式
- det ( A B ) = det ( B A ) = det ( A ) det ( B ) \det(AB) = \det(BA) = \det(A) \det(B) det(AB)=det(BA)=det(A)det(B)
- det ( A ) = ∏ i = 1 n λ i ( A ) \det(A) = \prod\limits_{i=1}^{n} \lambda_i(A) det(A)=i=1∏nλi(A)
- 迹
- t r ( A B ) = t r ( B A ) {\mathrm{tr}}(AB) = {\mathrm{tr}}(BA) tr(AB)=tr(BA)
- t r ( A ) = ∑ i = 1 n λ i ( A ) {\mathrm{tr}}(A) = \sum\limits_{i=1}^{n} \lambda_i(A) tr(A)=i=1∑nλi(A)
- 特征值
- A v = λ v Av = \lambda v Av=λv
- det ( A − λ I ) = 0 \det(A - \lambda I) = 0 det(A−λI)=0
- A ∈ M n × n ( R ) ⟹ ( det ( A − λ I ) = 0 ⟹ det ( A − λ ˉ I ) = 0 ) A \in M_{n \times n}(\reals) \implies (\det(A - \lambda I) = 0 \implies \det(A - \bar{\lambda} I) = 0) A∈Mn×n(R)⟹(det(A−λI)=0⟹det(A−λˉI)=0)
- det ( B − 1 A B − λ I ) = det ( A − λ I ) \det(B^{-1}AB - \lambda I) = \det(A - \lambda I) det(B−1AB−λI)=det(A−λI) (相似矩阵)
- λ ( A B ) = λ ( B A ) \lambda(AB) = \lambda(BA) λ(AB)=λ(BA)
- A B = A − 1 ( B A ) A AB = A^{-1} (BA) A AB=A−1(BA)A
- A V = V Λ AV = V \Lambda AV=VΛ
- 如果 A A A可对角化 A = V Λ V − 1 A = V \Lambda V^{-1} A=VΛV−1 (几何重数=代数重数) (特征向量线性无关张成全空间)
- A = V J V − 1 A = V J V^{-1} A=
- 如果 A A A可对角化 A = V Λ V − 1 A = V \Lambda V^{-1} A=VΛV−1 (几何重数=代数重数) (特征向量线性无关张成全空间)