3-2. 子空間
Define Subspace
(V, +, ⋅ ): vector space,
- W⊆V
- (W, +, ⋅ ): vector space
則稱 W 為 V 之子空間(subspace)
Theorem of 子空間的充要條件
已知 V: vector space, W⊆V, W≠∅, 則下列敘述等價:
- W 為 V 的子空間
- ∀u⃗, v⃗∈W, u⃗ + v⃗∈W
- ∀c∈F, ∀v⃗∈W, cv⃗∈W
∀c, d∈F, ∀u⃗, v⃗∈W, cu⃗ + dv⃗∈W
Corollary of 2. & 3.
∀ci∈F, vi∈W, i = 1, 2, ..., k, i=1∑kcivi∈W
Corollary of 4.
→ 在 W⊆V 和 W≠∅ 的條件下,只需驗證向量加法及純量積的封閉性即可證明 W 為 V 的子空間
Square of R 的三型子空間
原點: {(0, 0)} = { 0⃗ }: zero space → 點
∀ vector spaces, ∃{ 0⃗ }
過原點直線 → 線
若非直線,則不具純量積封閉性
x−y 平面: R2 → 面
V 為 V 的子空間
Theorem of 兩子空間交集
W1, W2 are subspaces of V, 則 W1∩W2 is a subspace of V
- ⊆
- ≠∅
- 封閉性
proof:
- W1⊆V 且 W2⊆V, 則 W1∩W2⊆V
- ∵0⃗∈W1∩W2, ∴W1∩W2≠∅
- ∀c, d∈F, ∀u⃗, v⃗∈W1∩W2,
∵u⃗, v⃗∈W1 且 W1 is a subspace of V, ∴cu⃗ + dv⃗ = W1
∵u⃗, v⃗∈W2 且 W2 is a subspace of V, ∴cu⃗ + dv⃗ = W2
∴cu⃗ + dv⃗ = W1∩W2
兩子空間聯集
W1, W2 are subspaces of V, W1∩W2 不一定為 V 的 subspace
反例:
V = R2 has two subsapces W1 and W2,
- W1 = { (x, 0) | x∈R },
x 軸
- W2 = { (0, y) | y∈R },
y 軸
but W1∪W2 is not a subspace of V.
ex. (1, 0), (0, 1) ∈W1∪W2, but (1, 0) + (0, 1) = (1, 1) ∉W1∪W2
Theorem of 兩子空間聯集
W1, W2 are subspaces of V, W1∩W2⇔W1⊆W2 or W2⊆W1
Define 和空間 (Sum Space)
W1, W2 are subspaces of V, W1 + W2 = { w1⃗ + w2⃗ | w1⃗∈W1, w2⃗∈W2 }, 稱為 W1, W2 之和空間(sum space)
Theorem of 和空間為 V 的子空間
W1, W2 are subspaces of V, 則 W1 + W2 is a subspace of V
proof:
- W1⊆V 且 W2⊆V, 則 W1 + W2⊆V
- ∵0⃗ = 0⃗ + 0⃗∈W1 + W2, ∴W1 + W2≠∅
- ∀c, d∈F, ∀u⃗, v⃗∈W1 + W2,
u⃗ = u1⃗ + u2⃗, u1⃗∈W1, u2⃗∈W2
v⃗ = v1⃗ + v2⃗, v1⃗∈W1, v2⃗∈W2
⇒cu⃗ + dv⃗ = c(u1⃗ + u2⃗) + d(v1⃗ + v2⃗) = (cu1⃗ + dv1⃗) + (cu2⃗ + dv2⃗) ∈W1 + W2
Define 四個基本子空間
A∈Fm×n,
- 核空間(kernel of A, nullspace, N(A)):
- ker(A) = { x⃗: n×1 | Ax⃗ = 0⃗ }
- 齊次解集,收集 Ax⃗ = 0⃗ 之 x⃗
- 行空間(column space):
- CS(A) = { Ax⃗: m×1 | x⃗: n×1 }
- 收集 Ax⃗ = y⃗ 之 y⃗
- 左核空間(left kernel of A, left nullspace):
- Lker(A) = { x⃗: 1×m | x⃗A = 0⃗ }
- 收集 x⃗A = 0⃗ 之 x⃗
- 列空間(row space):
- RS(A) = { x⃗A: 1×n | x⃗: 1×m }
- 收集 x⃗A = y⃗ 之 y⃗
ex.A=[112132]⎣⎡???⎦⎤=[00] ⇒ ker(A) =⎩⎨⎧⎣⎡aa−a⎦⎤∣a∈R⎭⎬⎫;
ex.B=[112200]⎣⎡???⎦⎤=[bb] ⇒ CS(B) ={[bb]∣b∈R}
Theorem of 四個基本子空間
A∈Fm×n,
- ker(A) is a subspace of Fn×1
- CS(A) is a subspace of Fm×1
∀y⃗∈CS(A) ⇔y⃗ = Ax⃗, for some x⃗
- Lker(A) is a subspace of F1×m
- RS(A) is a subspace of F1×n
proof:
- ∵A⋅0⃗ = 0⃗, ∴0⃗∈ker(A) ⇒ker(A) ≠∅
∀c, d∈F, ∀x1⃗, x2⃗∈ker(A) ⇒Ax1⃗ = Ax2⃗ = 0⃗
⇒A(cx1⃗ + dx2⃗) = cAx1⃗ + dAx2⃗ = c⋅0⃗ + d⋅0⃗ = 0⃗
⇒cx1⃗ + dx2⃗∈ker(A)
3 同理
- 0⃗ = A⋅0⃗∈CS(A) ⇒CS(A) ≠∅
∀c, d∈F, ∀y1⃗, y2⃗∈CS(A) ⇒y1⃗ = Ax1⃗, y2⃗ = Ax2⃗, x1⃗, x2⃗: n×1
⇒cy1⃗ + dy2⃗ = cAx1⃗ + dAx2⃗ = A(cx1⃗ + dx2⃗) ∈CS(A)
4 同理
四個基本子空間 with Invertible Matrix
- ker(B) ⊆ker(AB), 當 A is nonsingular 時, ker(B) = ker(AB)
ABx⃗ = 0⃗→Bx⃗ = 0⃗
- Lker(A) ⊆Lker(AB), 當 B is nonsingular 時, Lker(A) = Lker(AB)
- CS(AB) ⊆CS(A), 當 B 為可逆時, CS(AB) = CS(A)
- RS(AB) = RS(B)
四個基本子空間 with Row and Column Equivalence
A, B∈Fm×n,
- A 列等價於 B, 則
- ker(A) = ker(B)
- RS(A) = RS(B)
- A 行等價於 B, 則
- Lker(A) = Lker(B)
- CS(A) = CS(B)