3-2. 子空間

Define Subspace

(VV, +, \cdot ): vector space,

  • WVW \subseteq V

  • (WW, +, \cdot ): vector space

則稱 WWVV 之子空間(subspace)

Theorem of 子空間的充要條件

已知 VV: vector space, WVW \subseteq V, WW \ne \emptyset, 則下列敘述等價: 1. WWVV 的子空間 2. u\forall \vec{u}, vW\vec{v} \in W, u\vec{u} + vW\vec{v} \in W 3. cF\forall c \in F, vW\forall \vec{v} \in W, cvWc\vec{v} \in W 4. c\forall c, dFd \in F, u\forall \vec{u}, vW\vec{v} \in W, cuc\vec{u} + dvWd\vec{v} \in W

Corollary of 2. & 3.

  1. ciF\forall c_i \in F, viWv_i \in W, ii = 11, 22, ..., kk, i=1kciviW\displaystyle\sum_{i=1}^k c_i v_i \in W

    Corollary of 4.

\rightarrowWVW \subseteq VWW \ne \emptyset 的條件下,只需驗證向量加法純量積封閉性即可證明 WWVV 的子空間

Square of R 的三型子空間

  1. 原點: {(00, 00)} = { 0\vec{0} }: zero space \rightarrow

    \forall vector spaces, \exists{ 0\vec{0} }

  2. 過原點直線 \rightarrow

    若非直線,則不具純量積封閉性

  3. xyx-y 平面: R2R^2 \rightarrow

    VVVV 的子空間

Theorem of 兩子空間交集

W1W_1, W2W_2 are subspaces of VV, 則 W1W2W_1 \cap W_2 is a subspace of VV

  1. \subseteq

  2. \ne \emptyset

  3. 封閉性

proof:

  1. W1VW_1 \subseteq VW2VW_2 \subseteq V, 則 W1W2VW_1 \cap W_2 \subseteq V

  2. 0W1W2\because \vec{0} \in W_1 \cap W_2, W1W2\therefore W_1 \cap W_2 \ne \emptyset

  3. c\forall c, dFd \in F, u\forall \vec{u}, vW1W2\vec{v} \in W_1 \cap W_2,

    u\because \vec{u}, vW1\vec{v} \in W_1W1W_1 is a subspace of VV, cu\therefore c\vec{u} + dvd\vec{v} = W1W_1

    u\because \vec{u}, vW2\vec{v} \in W_2W2W_2 is a subspace of VV, cu\therefore c\vec{u} + dvd\vec{v} = W2W_2

    cu\therefore c\vec{u} + dvd\vec{v} = W1W2W_1 \cap W_2

兩子空間聯集

W1W_1, W2W_2 are subspaces of VV, W1W2W_1 \cap W_2 不一定VV 的 subspace

反例:

VV = R2R^2 has two subsapces W1W_1 and W2W_2,

  • W1W_1 = { (xx, 00) | xRx \in R },

    xx

  • W2W_2 = { (00, yy) | yRy \in R },

    yy

but W1W2W_1 \cup W_2 is not a subspace of VV.

ex.^{ex.} (11, 00), (00, 11) W1W2\in W_1 \cup W_2, but (11, 00) + (00, 11) = (11, 11) W1W2\notin W_1 \cup W_2

Theorem of 兩子空間聯集

W1W_1, W2W_2 are subspaces of VV, W1W2W1W2W_1 \cap W_2 \Leftrightarrow W_1 \subseteq W_2 or W2W1W_2 \subseteq W_1

Define 和空間 (Sum Space)

W1W_1, W2W_2 are subspaces of VV, W1W_1 + W2W_2 = { w1\vec{w_1} + w2\vec{w_2} | w1W1\vec{w_1} \in W_1, w2W2\vec{w_2} \in W_2 }, 稱為 W1W_1, W2W_2 之和空間(sum space)

Theorem of 和空間為 V 的子空間

W1W_1, W2W_2 are subspaces of VV, 則 W1W_1 + W2W_2 is a subspace of VV

proof:

  1. W1VW_1 \subseteq VW2VW_2 \subseteq V, 則 W1W_1 + W2VW_2 \subseteq V

  2. 0\because \vec{0} = 0\vec{0} + 0W1\vec{0} \in W_1 + W2W_2, W1\therefore W_1 + W2W_2 \ne \emptyset

  3. c\forall c, dFd \in F, u\forall \vec{u}, vW1\vec{v} \in W_1 + W2W_2,

    u\vec{u} = u1\vec{u_1} + u2\vec{u_2}, u1W1\vec{u_1} \in W_1, u2W2\vec{u_2} \in W_2

    v\vec{v} = v1\vec{v_1} + v2\vec{v_2}, v1W1\vec{v_1} \in W_1, v2W2\vec{v_2} \in W_2

    cu\Rightarrow c\vec{u} + dvd\vec{v} = cc(u1\vec{u_1} + u2\vec{u_2}) + dd(v1\vec{v_1} + v2\vec{v_2}) = (cu1c\vec{u_1} + dv1d\vec{v_1}) + (cu2c\vec{u_2} + dv2d\vec{v_2}) W1\in W_1 + W2W_2

Define 四個基本子空間

AFm×nA \in F^{m \times n},

  • 核空間(kernel of AA, nullspace, NN(AA)):

    • kerker(AA) = { x\vec{x}: n×1n \times 1 | AxA\vec{x} = 0\vec{0} }

    • 齊次解集,收集 AxA\vec{x} = 0\vec{0}x\vec{x}

  • 行空間(column space):

    • CSCS(AA) = { AxA\vec{x}: m×1m \times 1 | x\vec{x}: n×1n \times 1 }

    • 收集 AxA\vec{x} = y\vec{y}y\vec{y}

  • 左核空間(left kernel of AA, left nullspace):

    • LkerLker(AA) = { x\vec{x}: 1×m1 \times m | xA\vec{x}A = 0\vec{0} }

    • 收集 xA\vec{x}A = 0\vec{0}x\vec{x}

  • 列空間(row space):

    • RSRS(AA) = { xA\vec{x}A: 1×n1 \times n | x\vec{x}: 1×m1 \times m }

    • 收集 xA\vec{x}A = y\vec{y}y\vec{y}

ex.A=[123112][???]=[00]^{ex.} A = \begin{bmatrix} 1 & 2 & 3 \\ 1 & 1 & 2 \end{bmatrix} \begin{bmatrix} ? \\ ? \\ ? \end{bmatrix} = \begin{bmatrix} 0 \\ 0 \end{bmatrix} \Rightarrow kerker(AA) ={[aaa]aR}= \begin{Bmatrix} \begin{bmatrix} a \\ a \\ -a \end{bmatrix} | a \in R \end{Bmatrix};

ex.B=[120120][???]=[bb]^{ex.} B = \begin{bmatrix} 1 & 2 & 0 \\ 1 & 2 & 0 \end{bmatrix} \begin{bmatrix} ? \\ ? \\ ? \end{bmatrix} = \begin{bmatrix} b \\ b \end{bmatrix} \Rightarrow CSCS(BB) ={[bb]bR}= \begin{Bmatrix} \begin{bmatrix} b \\ b \end{bmatrix} | b \in R \end{Bmatrix}

Theorem of 四個基本子空間

AFm×nA \in F^{m \times n}, 1. kerker(AA) is a subspace of Fn×1F^{n \times 1} 2. CSCS(AA) is a subspace of Fm×1F^{m \times 1}

yCS\forall \vec{y} \in CS(AA) y\Leftrightarrow \vec{y} = AxA\vec{x}, for some x\vec{x} 3. LkerLker(AA) is a subspace of F1×mF^{1 \times m} 4. RSRS(AA) is a subspace of F1×nF^{1 \times n}

proof:

  1. A0\because A \cdot \vec{0} = 0\vec{0}, 0ker\therefore \vec{0} \in ker(AA) ker\Rightarrow ker(AA) \ne \emptyset

    c\forall c, dFd \in F, x1\forall \vec{x_1}, x2ker\vec{x_2} \in ker(AA) Ax1\Rightarrow A\vec{x_1} = Ax2A\vec{x_2} = 0\vec{0}

    A\Rightarrow A(cx1c\vec{x_1} + dx2d\vec{x_2}) = cAx1cA\vec{x_1} + dAx2dA\vec{x_2} = c0c \cdot \vec{0} + d0d \cdot \vec{0} = 0\vec{0}

    cx1\Rightarrow c\vec{x_1} + dx2kerd\vec{x_2} \in ker(AA)

    3 同理

  2. 0\vec{0} = A0CSA \cdot \vec{0} \in CS(AA) CS\Rightarrow CS(AA) \ne \emptyset

    c\forall c, dFd \in F, y1\forall \vec{y_1}, y2CS\vec{y_2} \in CS(AA) y1\Rightarrow \vec{y_1} = Ax1A\vec{x_1}, y2\vec{y_2} = Ax2A\vec{x_2}, x1\vec{x_1}, x2\vec{x_2}: n×1n \times 1

    cy1\Rightarrow c\vec{y_1} + dy2d\vec{y_2} = cAx1cA\vec{x_1} + dAx2dA\vec{x_2} = AA(cx1c\vec{x_1} + dx2d\vec{x_2}) CS\in CS(AA)

    4 同理

四個基本子空間 with Invertible Matrix

  1. kerker(BB) ker\subseteq ker(ABAB), 當 AA is nonsingular 時, kerker(BB) = kerker(ABAB)

    ABxAB\vec{x} = 0Bx\vec{0} \rightarrow B\vec{x} = 0\vec{0}

  2. LkerLker(AA) Lker\subseteq Lker(ABAB), 當 BB is nonsingular 時, LkerLker(AA) = LkerLker(ABAB)

  3. CSCS(ABAB) CS\subseteq CS(AA), 當 BB 為可逆時, CSCS(ABAB) = CSCS(AA)

  4. RSRS(ABAB) = RSRS(BB)

四個基本子空間 with Row and Column Equivalence

AA, BFm×nB \in F^{m \times n}, 1. AA 列等價BB, 則

  • kerker(AA) = kerker(BB)

  • RSRS(AA) = RSRS(BB)

    1. AA 行等價BB, 則

  • LkerLker(AA) = LkerLker(BB)

  • CSCS(AA) = CSCS(BB)

Last updated

Was this helpful?