Chapter 3 Orthogonality
3.1 Orthogonal Vectors and Subspaces
the length
∣
∣
x
∣
∣
||x||
∣∣x∣∣ in
R
n
R^n
Rn is the positive square root of
x
T
x
x^Tx
xTx:
Length square
∣
∣
x
∣
∣
2
=
x
1
2
+
x
2
2
+
⋯
+
x
n
2
=
x
T
x
(1)
\text{Length square } \ \ \ \ \ \ ||x||^2 = x_1^2 + x_2^2 + \dots + x_n^2 = x^Tx \tag1
Length square ∣∣x∣∣2=x12+x22+⋯+xn2=xTx(1)
Orthogonal Vectors
Sides of a right triangule ∣ ∣ x ∣ ∣ 2 + ∣ ∣ y ∣ ∣ 2 = ∣ ∣ x − y ∣ ∣ 2 (2) \text{Sides of a right triangule} \ \ \ \ \ ||x||^2 + ||y||^2 = ||x-y||^2 \tag 2 Sides of a right triangule ∣∣x∣∣2+∣∣y∣∣2=∣∣x−y∣∣2(2)
Orthogonal vectors x T y = x 1 y 1 + ⋯ + x n y n = 0 (3) \text{Orthogonal vectors} \ \ \ \ \ x^Ty = x_1y_1 + \dots + x_ny_n = 0 \tag3 Orthogonal vectors xTy=x1y1+⋯+xnyn=0(3)
Inner product x T y = [ x 1 … x n ] [ y 1 ⋮ y n ] = x 1 y 1 + ⋯ + x n y n (4) \text{Inner product }\ \ \ \ \ x^Ty = \left[ \begin{matrix} x_1 & \dots & x_n \end{matrix}\right] \left[ \begin{matrix} y_1 \\ \vdots \\ y_n \end{matrix}\right] = x_1y_1 + \cdots + x_ny_n \tag4 Inner product xTy=[x1…xn]⎣⎢⎡y1⋮yn⎦⎥⎤=x1y1+⋯+xnyn(4)
3A The inner product $x^Ty $ is zero if and only if x x x and y y y are orthogonal vectors, If x T y > 0 x^Ty > 0 xTy>0, their angle is less than 9 0 ∘ 90^\circ 90∘, If x T y < 0 x^Ty < 0 xTy<0, their angle is greater than 9 0 ∘ 90^\circ 90∘.
If nonzero vector v 1 , … v k v_1, \dots v_k v1,…vk are mutually orthogonal (every vector is perpendicular to every other), then those vectors are linearly independent.
Orthogonal Subspaces
3B Two subspaces V V V and W W W of the same space R n R^n Rn are orthogonal if every vector v v v in V V V is orthogonal to every vector w w w in W W W: v T w = 0 v^Tw = 0 vTw=0 for all v v v and w w w.
3C Fundamental theorem of orthogonality The row space is orthogonal to the nullspace (in R n R^n Rn). The column space is orthogonal to the left nullspace (in R m R^m Rm).
**Definition. Given a subspace V V V of R n R^n Rn, the space of all vectors orthogonal to V V V is called the orthogonal complement of V V V. } It is denoted by V ⊥ = " V p e r p " V^{\perp}= "V perp" V⊥="Vperp" **
3D Fundamental Theorem of Linear Algebra, Part II
The nullspace is the orthogonal complement of the row space in R n R^n Rn.
The left nullspace if the orthogonal complement of the column space in R m R^m Rm.
3E A x = b Ax=b Ax=b is solvable if and only if y T b = 0 y^Tb = 0 yTb=0 whenever y T A = 0 y^TA = 0 yTA=0
The Matrix and the Subspaces
If W = V ⊥ W = V^{\perp} W=V⊥ then V = W ⊥ V=W^{\perp} V=W⊥ and d i m V + d i m W = n dimV + dimW = n dimV+dimW=n.
In other words V ⊥ ⊥ = V V^{\perp\perp} = V V⊥⊥=V. The dimensions of V V V and W W W are right, and the whole space R n R^n Rn is being decomposed into two perpendicular parts.
When multiplied by A A A, this is A x = A x r + A x n Ax = Ax_r + Ax_n Ax=Axr+Axn, :
x n x_n xn is the nullspace component, A x n = 0 Ax_n = 0 Axn=0
x r x_r xr is the row space component, A x r = A x Ax_r = Ax Axr=Ax.
3F From the row space to column space, A A A is actually invertible. Every vector b b b in the column space comes from exactly one vector x r x_r xr in the row space.
Every matrix transforms its row space onto its column space.
3.2 Cosines and Projections onto lines
The line connecting b b b to p p p is perpendicular to a a a.
The point p p p is the projection of b b b onto the subspace.