Yejun Cheon
Yejun Cheon
log
study
read
coffee
Sign In
4๏ธโƒฃ

Lecutre 04

Intro

Linear Combination

๐‘1 ๐ฏ๐Ÿ + ๐‘2 ๐ฏ๐Ÿ + โ‹ฏ + ๐‘๐‘ ๐ฏ๐’‘
๐‘“๐‘œ๐‘Ÿ ๐‘ ๐‘๐‘Ž๐‘™๐‘Ž๐‘Ÿ๐‘  ๐‘๐‘– ๐‘– = 1,2, โ€ฆ p
์ด๊ฑธ matrix form ์œผ๋กœ ์ ์œผ๋ฉด

Subspace

์–ด๋–ค ์ง‘ํ•ฉ ๋‚ด์˜ ์›์†Œ๋“ค์˜ ์„ ํ˜•๊ฒฐํ•ฉ๋„ ์ง‘ํ•ฉ์•ˆ์— ์žˆ๋Š” ๊ฒฝ์šฐ. โ†’ 1v +(-1)v = 0 ๋„ ํฌํ•จํ•ด์•ผํ•˜๋‹ˆ๊นŒ ์˜๋ฒกํ„ฐ ํฌํ•จ๋„ ํ•„์ˆ˜์ž„.

n-dim ์—์„œ subspace์˜ ์ข…๋ฅ˜

2์ฐจ์›์—์„œ
โ€ข
zero space
โ€ข
line through the origin
โ€ข
โ„^2
3์ฐจ์›์—์„œ
โ€ข
zero space
โ€ข
line through the origin
โ€ข
plane through the origin
โ€ข
โ„^3

Linearly (in)dependence

Linearly dependent (rank < #(var) )ํ•˜๋ฉด Ax=0์ด ๋ฌด์ˆ˜ํžˆ ๋งŽ์€ ํ•ด๋ฅผ ๊ฐ€์ง.
Linearly independent (rank = #(var))ํ•˜๋ฉด Ax=0์ด x=0(trivial solution)์œผ๋กœ ์œ ์ผํ•œ ํ•ด๋ฅผ ๊ฐ€์ง.
why? ์„ ํ˜•์ข…์†์˜ ์ •์˜ ๐‘1๐ฏ1 + ๐‘2 ๐ฏ2 + โ‹ฏ + ๐‘๐‘ ๐ฏp = ๐ŸŽ ๋ฅผ ์ƒ๊ฐํ•˜๋ฉด ๋‹น์—ฐํ•จ.

Row Operation์€ column space๋ฅผ ๋ณด์กดํ•˜์ง€ ์•Š๋Š”๋‹ค.

๊ธฐ์ € ์ฐพ๊ธฐ, ์„ ํ˜• ๋…๋ฆฝ์„ฑ ํŒ๋ณ„, Column space ํ‘œํ˜„ํ•˜๊ธฐ ํ• ๋•Œ ์œ ์˜์‚ฌํ•ญ
REF์˜ pivot column index๋ฅผ ๊ณ ๋ฅด๋˜, ์‹ค์ œ๋กœ๋Š” ์›๋ž˜ ํ–‰๋ ฌ์˜ ํ•ด๋‹น column์„ ์„ ํƒํ•ด์•ผํ•œ๋‹ค.
Pivot์€ ์œ„์น˜๋งŒ, Column์€ ์›๋ณธ์—์„œ

span๋˜๋Š” ๋ฒกํ„ฐ์ธ์ง€ ํŒ๋ณ„๊ณผ ๋™์น˜์ธ ๋ช…์ œ

์–ด๋–ค ๋ฒกํ„ฐ u ๊ฐ€ u_1, u_2์— ์˜ํ•ด span ๋˜๋Š”์ง€ ํŒ๋ณ„ํ•˜๋Š” ๋ฌธ์ œ์™€ ๋™์น˜์ธ ๋ฌธ์ œ๋“ค.

\exists x \ \text{s.t.} \ A x = \mathbf{u}

\text{where } A = [\mathbf{u}_1 \ \mathbf{u}_2]
Ax=u ์˜ ํ•ด x๊ฐ€ ์กด์žฌํ•œ๋‹ค.

\Longleftrightarrow \ \mathbf{u} \in \mathcal{C}(A)
A์˜ column์ธ u_1 u_2์˜ ์„ ํ˜•๊ฒฐํ•ฉ์œผ๋กœ ์ƒ๊ธฐ๋Š” ๊ณต๊ฐ„์ธ C(A)์— u๊ฐ€ ํฌํ•จ๋œ๋‹ค.

\Longleftrightarrow \ \operatorname{rank}(A) \stackrel{?}{=} \operatorname{rank}([A \mid \mathbf{u}])
A ์˜†์— u ์ปฌ๋Ÿผ์„ ์ด์–ด ๋ถ™์—ฌ ์ถ”๊ฐ€ํ•œ ์ฆ๊ฐ•ํ–‰๋ ฌ์˜ rank = dim(ColSpace) ๊ฐ€ ๋™์ผํ•˜๋‹ค.
์ƒˆ๋กœ์šด ์ปฌ๋Ÿผ์ด ๋‹ค๋ฅธ ์ปฌ๋Ÿผ๋“ค์˜ ์„ ํ˜•๊ฒฐํ•ฉ์œผ๋กœ ํ‘œํ˜„๋œ๋‹ค.

Null Space / Column space

kernel, image

ํŒŒ๋ž€์„ ์€ null space = kernel by A โ†’ ์˜๋ฒกํ„ฐ๋กœ ๊ฐ„๋‹ค.
๊ฒ€์ •์„ ์€ ์ž„์˜์˜ x=[ a b]^T๋ฅผ ์žก์•˜์„ ๋•Œ, [a-2b a-2b]^T ์ฆ‰ y=x ์ง์„ ์œผ๋กœ ๊ฐ€๋Š”๊ฑธ ๋‚˜ํƒ€๋ƒ„
๋ฌธ์ œํ’€์ด
์–ด๋–ค ์ง์„ ์— ๋Œ€ํ•˜์—ฌ linear transformation ํ•œ image๋ฅผ ๊ตฌํ•˜๋Š” ๋ฐฉ๋ฒ•.
์ง์„ ์„ ๋ฒกํ„ฐํ‘œ๊ธฐ(direction vector t + passing vector)๋กœ ๋ฐ”๊พผ๋‹ค.
matrix multiplication์„ ์ง„ํ–‰ํ•œ๋‹ค.
๋‹ค์‹œ ๋ฐฉ์ •์‹ ํ‘œ๊ธฐ๋กœ ๋ฐ”๊พผ๋‹ค.

determinant์˜ ๊ธฐํ•˜์  ์˜๋ฏธ

det์˜ ํฌ๊ธฐ๊ฐ€ ํ–‰๋ ฌ์˜ column๋“ค๋กœ ๊ตฌ์„ฑ๋œ ๋„ํ˜•์˜ ๋„“์ด, ๋ถ€ํ”ผ์˜ ๋ณ€ํ™”์˜ scale factor์ž„

Orthogonal matrix

definition

QQ^T=I ์ธ n x n matrix

properties

1.
orthonormal columns
a_i \cdot a_j๊ฐ€ Kronecker delta ์„ฑ์งˆ์„ ๊ฐ€์ง€๊ณ  ์žˆ์–ด์•ผ ๊ณฑํ–ˆ์„ ๋•Œ Identity Matrix๊ฐ€ ๋œ๋‹ค. ๋‹ค์‹œ๋งํ•˜๋ฉด ์ž์‹ ๋ผ๋ฆฌ dot product๋ฅผ ํ–ˆ์„ ๋•Œ๋Š” 1์ด ๋‚˜์™€์•ผํ•˜๊ณ  = norm ์ด 1์ด๋‹ค.
๋‹ค๋ฅธ column ๊ณผ dot product์„ ํ–ˆ์„ ๋•Œ๋Š” 0์ด ๋‚˜์™€์•ผํ•œ๋‹ค= ๋‘ ์ปฌ๋Ÿผ์€ ์ง๊ตํ•œ๋‹ค.
1.
det(A) = +-1
det(AA^T)= det(A)det(A^T) = det(A)^2 = det(I) = 1
determinant์˜ ์„ฑ์งˆ์— ์˜ํ•ด ๋‹ค์Œ๊ณผ ๊ฐ™์ด ๋œ๋‹ค.
1.
dot product preserved ๋œ๋‹ค. โ†” norm, angle, distance๊ฐ€ preserved ๋œ๋‹ค.
u \cdot v = u^T v ์„ ์‚ฌ์šฉ
dot product preserve ๋˜๋Š” ์ 











\begin{aligned}
(Q\mathbf{x}) \cdot (Q\mathbf{y})
&= (Q\mathbf{x})^T (Q\mathbf{y}) \\
&= \mathbf{x}^T Q^T Q \mathbf{y} \\
&= \mathbf{x}^T I \mathbf{y} \\
&= \mathbf{x}^T \mathbf{y} \\
&= \mathbf{x} \cdot \mathbf{y}
\end{aligned}

norm์ด ๋ณด์กด๋œ๋‹ค.








\|Qx\| = \sqrt{(Qx) \cdot (Qx)} = \sqrt{x \cdot x} = \|x\|
๊ฐ๋„๊ฐ€ ๋ณด์กด๋œ๋‹ค.








\cos\theta = \frac{x \cdot y}{\|x\|\|y\|}

\cos\thetaโ€™ = \frac{(Qx) \cdot (Qy)}{\|Qx\|\|Qy\|}
= \frac{x \cdot y}{\|x\|\|y\|} = \cos\theta

Linear Transformation

Reflection

Rotation

constant matrix
โ€ข
det(R_\theta) = 1
โ€ข
orthogonal matrix(์ปฌ๋Ÿผ ๊ฐ„ ์ง๊ต, norm ์ด 1)
scaling factor ๊ฐ€ ๊ณฑํ•ด์ง„ ๋ฒ„์ ผ๋„ ์žˆ์Œ. ๊ฐ๋„ ๋ณด์กด, ์Šค์ผ€์ผ๋ง.

Reflection about y = tan \theta x

ํ•ต์‹ฌ ์•„์ด๋””์–ด : tan x ์„ ์„ x์ถ•์œผ๋กœ ํšŒ์ „์‹œํ‚ค๊ณ , x์ถ•์„ ๊ธฐ์ค€์œผ๋กœ reflection ํ•œ๋’ค์—, 2theta๋งŒํผ ์›๋ณต์‹œํ‚จ๋‹ค.








A =
\begin{bmatrix}
\cosฮธ & -\sinฮธ \\
\sinฮธ & \cosฮธ
\end{bmatrix}
\begin{bmatrix}
1 & 0 \\
0 & -1
\end{bmatrix}
\begin{bmatrix}
\cosฮธ & \sinฮธ \\
-\sinฮธ & \cosฮธ
\end{bmatrix}








A =
\begin{bmatrix}
\cos^2ฮธ - \sin^2ฮธ & 2\sinฮธ\cosฮธ \\
2\sinฮธ\cosฮธ & \sin^2ฮธ - \cos^2ฮธ
\end{bmatrix}
โ€ข
det(R_\theta) = -1
โ€ข
orthogonal matrix(์ปฌ๋Ÿผ ๊ฐ„ ์ง๊ต, norm ์ด 1)

Quiz2

PartA

Q1.

A \in \mathbb{R}^{m \times n} : \mathbb{R}^{n} \to \mathbb{R}^{m}์ด๋‹ค.
dim(C(A)) = rank(A)
dim(codomain) = m
dim(N(A)) = dim(domain) - rank(A)

Q2. rank(A) = rank(A|b) ์ธ ๊ฒฝ์šฐ์— Ax=b์˜ solution์ด ์กด์žฌํ•จ์„ ์ฆ๋ช…ํ•˜๊ธฐ

rank(A) = rank(A|b)ย  ๋Š” dim(C(A)) = dim(C(A|b)) ์™€ ๋™์น˜์ž„.
๋”ฐ๋ผ์„œ ์ฆ๊ฐ•ํ–‰๋ ฌ์˜ ์ถ”๊ฐ€๋œ ์ปฌ๋Ÿผ b๋Š” ์ด๋ฏธ Column space์— ํฌํ•จ๋œ ๊ฑฐ์ž„. b \in C(A)
\Leftrightarrow \exists x_i \in \mathbb{R} \text{ s.t. } b = \sum_{i=1}^{n} x_i a_i ๊ทธ ์–˜๊ธฐ๋Š”b๋ฅผ linear combination์œผ๋กœ ์“ธ ์ˆ˜ ์žˆ๋‹ค๋Š”๊ฑฐ๊ณ 
\Leftrightarrow \exists x \in \mathbb{R}^n \text{ s.t. } Ax = b ํ–‰๋ ฌ๊ณฑ ํ˜•ํƒœ๋กœ ํ‘œํ˜„ํ•˜๋ฉด Ax=b์˜ solution์ด ์กด์žฌํ•œ๋‹ค๋Š” ๊ฑฐ์ž„.

PartB

Q2.

T_{A} : \mathbb{R} ^2 \to \mathbb{R}^3, such that \begin{cases}
T\!\left(\begin{bmatrix}1\\[2pt]1\end{bmatrix}\right)=\begin{bmatrix}2\\[2pt]1\\[2pt]-1\end{bmatrix},\\[8pt]
T\!\left(\begin{bmatrix}2\\[2pt]3\end{bmatrix}\right)=\begin{bmatrix}4\\[2pt]0\\[2pt]2\end{bmatrix},
\end{cases}ย 
evaluate T\!\left(\begin{bmatrix}3\\[2pt]-1\end{bmatrix}\right).ย 

Sol1)

transtition matrix A๋Š” \mathbb{R}^{3 \times 2} ์˜ ๊ตฌ์กฐ์ž„. codomain์˜ dimension ์ด ์•ž, domain์˜ dimension์ด ๋’ค์ชฝ.
Find c_1, c_2 \in \mathbb{R} s.t. \begin{bmatrix}
3\\-1
\end{bmatrix}
=c_1
\begin{bmatrix}
1\\1
\end{bmatrix}
+c_2
\begin{bmatrix}
2\\3
\end{bmatrix}

.
linear combination์œผ๋กœ [3 -1]^T ๋ฅผ ํ‘œํ˜„ํ•˜๊ณ  ์ด๊ฑธ ๋งŒ์กฑํ•˜๋Š” c_1, c_2๋ฅผ ์ฐพ๋Š”๋‹ค.
Linear Transformation ์˜ ์„ฑ์งˆ(linearity)์— ์˜ํ•ด์„œ T(\begin{bmatrix}
3\\-1
\end{bmatrix})
=c_1
T(\begin{bmatrix}
1\\1
\end{bmatrix})
+c_2
T(\begin{bmatrix}
2\\3
\end{bmatrix})
์˜ ์‹์ด ์œ ์ง€๋จ.
์ฐพ์€ c_1๊ณผ c_2๋ฅผ ์ ์šฉํ•˜์—ฌ ๊ตฌํ•˜๊ณ ์ž ํ•˜๋Š” ๊ฐ’์„ ๊ตฌํ•œ๋‹ค.

Q2.2 A๋ฅผ ๊ตฌํ•˜์‹œ์˜ค

Transition matrix A๋ฅผ ์ฐพ์•„๋ผ.์™€ ๊ฐ™์€ ์ƒˆ๋ผ ๋ฌธ์ œ๊ฐ€ ์žˆ๋‹ค๋ฉด ์ด ์ „์ดํ–‰๋ ฌ์„ ๋จผ์ € ๊ตฌํ•˜๊ณ  ๊ทธ ํ–‰๋ ฌ์˜ ๊ณฑ์œผ๋กœ Linear Transformation์„ ์ ์šฉํ•˜๋Š” ๊ฒƒ๋„ ์ข‹์Œ.
A\begin{bmatrix}
1 & 2 \\
1 &3
\end{bmatrix}
=
\begin{bmatrix}
2 &4\\
1 &0 \\
-1 &2
\end{bmatrix}

๋‘ ๊ฐœ์˜ ๋ฒกํ„ฐ์˜ ๊ณฑ ๊ฒฐ๊ณผ๋ฅผ concat ํ•˜์—ฌ ํ–‰๋ ฌ๊ณฑ์„ ๊ตฌ์„ฑ. Square matrix์ด๊ธฐ์— ์—ญํ–‰๋ ฌ๋กœ ์–‘๋ณ€์— ๊ณฑํ•ด์„œ A๋ฅผ ๊ตฌํ•  ์ˆ˜ ์žˆ์Œ.

Q2.3 range(T) ๋ฅผ ๊ตฌํ•˜์‹œ์˜ค

=Column Space of A = C(A)
rank(A) = 2 ์ด๊ธฐ ๋•Œ๋ฌธ์— Column space๊ฐ€ ์›์  ์ง€๋‚˜๋Š” ํ‰๋ฉด์ด๋‹ค.
์ด์ œ๋Š” ๊ตฌ์ฒด์ ์ธ ๋ฐฉ์ •์‹์„ ๊ตฌํ•ด์•ผํ•œ๋‹ค. normal vector๋ฅผ ๊ตฌํ•˜๋Š” ๋ฐฉ๋ฒ•. cross productํ•˜๋ฉด normal vector ๋‚˜์˜จ๋‹ค.
column space : 2x-8y-4z =0 ์ •๋ฆฌํ•˜๋ฉด x-4y-2z=0
image ๊ณต๊ฐ„์ด ์–ด๋””์— ๊ทธ๋ ค์ง€๋Š”์ง€ ํ‘œํ˜„ํ•  ์ˆ˜ ์žˆ์–ด์•ผํ•จ.
์ด ๊ฒฝ์šฐ์—๋Š” ์›์ ์„ ์ง€๋‚˜๋Š” ํ‰๋ฉด x-4y-z=0 ์œ„์— ์žˆ๋Š” ํƒ€์›์ž„. ์žฅ์ถ•์˜ ๊ธธ์ด๋Š” max( Ax ), ๋‹จ์ถ•์˜ ๊ธธ์ด๋Š” min( Ax )

Part3

Q1

T(x,y,z) = (2x-y,x+z,y-z)

show that it is linear

ํ–‰๋ ฌ์—๋‹ค๊ฐ€ [x,y,z]^T๋ฅผ ๊ณฑํ•œ๊ฑฐ๋กœ ํ‘œํ˜„ํ•˜๋ฉด ๋จ. T(x) = Ax์ธ ํ–‰๋ ฌ A๋งŒ ์ฐพ์œผ๋ฉด ์„ ํ˜•์„ฑ์ด ์ฆ๋ช…๋จ. (image๊ฐ€ ํ–‰๋ ฌ๊ณฑ ํ˜•ํƒœ๋กœ ํ‘œํ˜„๋จ์„ ๋ณด์ด๋ฉด ๋จ.)

is w = [3\ 3\ 0]^T in the range of T?

w\in C(A)์ธ์ง€๋ฅผ ๋ฌผ์–ด๋ณด๋Š” ๋ฌธ์ œ.

Intro ํ›„๋ฐ˜๋ถ€

Orthogonal matrix

A^TA = I_n์ธ A
(A^TA)_{ij} = a_i \cdot a_j (where a_i = i th col of A)

property

1.
a_i \cdot a_j = \delta_{ij}= \begin{cases}
0 & \text{if } i \neq j \\
1 & \text{if } i =j
\end{cases}
์ฆ‰, orthonormal column์„ ๊ฐ€์ง„๋‹ค.
2.
det(A^TA) = det(A^T)det(A) = det(A)^2= det(I) = 1 ์ด๋ฏ€๋กœ det(A) = 1 or -1์ด๋‹ค.
3.
x_1, x_2 \in \mathbb{R}^n์˜ ์ด๋ฏธ์ง€ ๋ฒกํ„ฐ์˜ ๋‚ด์ ์„ ์ •๋ฆฌํ•˜๋ฉดAx_1 \cdot Ax_2 = (Ax_1)^T(Ax_2) = x_1A^TAx_2 = x_1 \cdot x_2 ์›๋ž˜ ๊ฐ’์˜ ๋‚ด์ ๊ณผ ๋™์ผํ•˜๋‹ค. dot-product is preserved! โ†”norm& angle& distance๊ฐ€ ๋ชจ๋‘ preserved ๋จ.
4. A^{-1} = A^T

ex1.

show that A๊ฐ€ orthogonal์ž„์„ ๋ณด์—ฌ๋ผ.
๊ฐ ์ปฌ๋Ÿผ์˜ ํฌ๊ธฐ๊ฐ€ 1์ด๊ณ , ๋‘ ์ปฌ๋Ÿผ์„ ๋‚ด์ ํ•˜๋ฉด 0 ์ž„. ๊ทธ๋ž˜์„œ orthogonal์ž„์„ ๋ฐ”๋กœ ์•Œ ์ˆ˜ ์žˆ์Œ.
A = \begin{bmatrix}
cos\theta& -sin\theta \\
sin\theta& cos\theta \\
\end{bmatrix}
det(A) = c^2 + s^2 = 1 counter-clockwise,\theta-rotation matrix(angle, norm, distance ์œ ์ง€)
find inverse matrix of A.

ex2.

orthogonal์ž„์œผ๋กœ ๋ณด์—ฌ๋ผ.
definition์œผ๋กœ ํ‘ธ๋Š” ๋ฒ•.A^TA = I_n์ž„์„ ๋ณด์ธ๋‹ค. โ†’ ์—„์ฒญ ์˜ค๋ž˜๊ฑธ๋ฆผ.
property 1 ์„ ์‚ฌ์šฉํ•˜์—ฌ column๊ฐ„์˜ dot product๊ฐ€ 0์ธ์ง€, ๊ฐ column์˜ ํฌ๊ธฐ๊ฐ€ 1์ธ์ง€ ๋ณด์ด๋ฉด ํŽธํ•จ.
inverse function ๊ตฌํ•˜๋Š” ๋ฒ•. transpose ๊ตฌํ•˜๋ฉด ๋จ. by property 4์— ์˜ํ•ด.
det(A) ๊ตฌํ•˜๋Š” ๋ฒ•. 1 ์•„๋‹ˆ๋ฉด -1์ž„.
A=[v_1 \ v_2 \ v_3] \to det(A) = v_1 \cdot (v_2 \times v_3)
cross productํ•ด์„œ ๋ฐฉํ–ฅ์ด ๊ฐ™์œผ๋ฉด ์–‘์ˆ˜, ๋‹ค๋ฅด๋ฉด ์Œ์ˆ˜.

Online class

Orthogonal projection about line y=(tan \theta )x

v๋Š” norm์ด 1์ธ unit vector.
projection์˜ ๊ณต์‹ \text{proj}_{\mathbf{v}}(\mathbf{x}) = \frac{\mathbf{x} \cdot \mathbf{v}}{\mathbf{v} \cdot \mathbf{v}} \, \mathbf{v} ์ธ๋ฐ, v๊ฐ€ unit vector์—ฌ์„œ T(x) = (x \cdot v)v์ž„. ์—ฌ๊ธฐ์„œ ์•ž์— dot product๋กœ ๋˜์–ด์žˆ๋Š” ๋ถ€๋ถ„์ด ์Šค์นผ๋ผ๊ณ , v๊ฐ€ ๋ฐฉํ–ฅ์„ ๋‚˜ํƒ€๋ƒ„.
projection ๊ณต์‹ ์œ ๋„
๋ฒกํ„ฐ \mathbf{x}๋ฅผ \mathbf{v} ๋ฐฉํ–ฅ์œผ๋กœ ํˆฌ์˜ํ•œ ๋ฒกํ„ฐ๋Š” \mathbf{v}์™€ ๊ฐ™์€ ๋ฐฉํ–ฅ์„ ๊ฐ€์ง€๋˜ ํฌ๊ธฐ๋งŒ ๋‹ค๋ฅธ ํ˜•ํƒœ c\mathbf{v}๋กœ ํ‘œํ˜„๋˜๋ฉฐ, ์ˆ˜์ง ์กฐ๊ฑด (\mathbf{x} - c\mathbf{v}) \cdot \mathbf{v} = 0์œผ๋กœ๋ถ€ํ„ฐ c = \frac{\mathbf{x}\cdot \mathbf{v}}{\mathbf{v}\cdot \mathbf{v}}๊ฐ€ ๋œ๋‹ค. ๋”ฐ๋ผ์„œ ํˆฌ์˜ ๋ฒกํ„ฐ๋Š” \text{proj}{\mathbf{v}}(\mathbf{x}) = \frac{\mathbf{x}\cdot \mathbf{v}}{\mathbf{v}\cdot \mathbf{v}}\mathbf{v} ๋กœ ์ฃผ์–ด์ง„๋‹ค. ์ด๋Š” \mathbf{x}๊ฐ€ \mathbf{v} ๋ฐฉํ–ฅ์œผ๋กœ ์–ผ๋งˆ๋‚˜ ๊ฒน์น˜๋Š”์ง€๋ฅผ ๋‚˜ํƒ€๋‚ด๋Š” ๋‚ด์  \mathbf{x}\cdot\mathbf{v}๋ฅผ \mathbf{v}์˜ ํฌ๊ธฐ ์ œ๊ณฑ์œผ๋กœ ์ •๊ทœํ™”ํ•œ ๋’ค, ๊ทธ๋งŒํผ \mathbf{v} ๋ฐฉํ–ฅ์œผ๋กœ ๋Š˜๋ฆฐ ๋ฒกํ„ฐ์ด๋‹ค. ํŠนํžˆ \mathbf{v}๊ฐ€ ๋‹จ์œ„๋ฒกํ„ฐ๋ผ๋ฉด \mathbf{v}\cdot\mathbf{v}=1์ด ๋˜์–ด \text{proj}{\mathbf{v}}(\mathbf{x}) = (\mathbf{x}\cdot\mathbf{v})\mathbf{v}๋กœ ๋‹จ์ˆœํ™”๋œ๋‹ค.
๊ทผ๋ฐ ์ด dot product๋ฅผ ํ–‰๋ ฌํ˜•ํƒœ๋กœ ํ‘œํ˜„ํ•˜๋ฉด T(x) =(vv^T)xํ˜•ํƒœ๊ฐ€ ๋˜๊ณ , P=vv^T๋ฅผ Projection matrix๋ผ๊ณ  ํ•œ๋‹ค.
โ†’ ๊ฒฐ๋ก  : ์–ด๋–ค normal ๋ฒกํ„ฐ v์— projection ํ•˜๊ณ  ์‹ถ์œผ๋ฉด Projection Matrix P=vv^T๋ฅผ ์›ํ•˜๋Š” ๋ฒกํ„ฐ x์— ๊ณฑํ•˜์ž.
๋‹ค๋ฅธ ๋ฐฉ๋ฒ•์œผ๋กœ๋„ ์ด P๋ฅผ ๊ตฌํ•  ์ˆ˜๋„ ์žˆ๋Š”๋ฐ x๋ฅผ tan ์„ธํƒ€์— ๋Œ€ํ•ด reflectionํ•œ ๋ฒกํ„ฐ์™€ ํ‰๊ท ์„ ์ทจํ•ด projection์„ ํ‘œํ˜„ํ•˜๋Š”๊ฑฐ๋‹ค. cos/sin 2๋ฐฐ๊ฐ ๊ณต์‹์„ ์จ๋ณด๋ฉด ๋™์ผํ•จ์„ ์•Œ ์ˆ˜ ์žˆ๋‹ค.
๊ณต๊ฐ„์—์„œ์˜ ์˜๋ฏธ๊ฐ€ ํ—ท๊ฐˆ๋ ค ์ •๋ฆฌ๋ฅผ ํ•ด๋ณด์ž๋ฉด, projection vector๋Š” x๋ฒกํ„ฐ์—์„œ y=tan x ์ง์„ ์— ์ˆ˜์„ ์„ ๋‚ด๋ฆฐ ์ง๊ฐ ์‚ผ๊ฐํ˜•์˜ ๋ฐ‘๋ณ€ ๋ถ€๋ถ„์— ํ•ด๋‹นํ•˜๋Š” ๋ฒกํ„ฐ์ด๋‹ค. ๊ทธ๋ฆฌ๊ณ  P๋Š” x์—๋‹ค๊ฐ€ ๊ณฑํ•ด์ฃผ๊ธฐ๋งŒ ํ•œ๋‹ค๋ฉด ๋ฐ”๋กœ ๊ทธ projection vector๋ฅผ ๊ตฌํ•ด์ฃผ๋Š” ์„ ํ˜• ๋ณ€ํ™˜์„ ํ•ด์ฃผ๋Š” ์ „์ดํ–‰๋ ฌ์ด๋‹ค.
det(P) =0 : ํ•œ ์ ์„ ์••์ถ•๋˜๋‹ˆ๊นŒ det์ด 0์ž„
rank(P) = 1 : ์ถœ๋ ฅ์ด projection vector ์ง์„  1์ฐจ์› ๊ณต๊ฐ„์— ์กด์žฌํ•จ.
C(P) : y = (tan \theta)x : projection vector๋Š” ๋‹น์—ฐํžˆ ์ € ์ง์„  ์œ„์— ์ˆ˜์„ ์„ ๋‚ด๋ฆฌ๊ธฐ์— ์ € ์œ„์— ์กด์žฌํ•  ์ˆ˜ ๋ฐ–์— ์—†๋‹ค.
N(P) : y = (-cot \theta)x : ์›์ ์„ ์ง€๋‚˜๋ฉด์„œ tan x ์— ์ˆ˜์ง์ธ ์ด ์œ„์˜ ๋ฒกํ„ฐ x๋“ค์€ y = (tan \theta)x์— ์ˆ˜์„ ์„ ๋‚ด๋ฆฌ๋ฉด ์›์ ์ด๋‹ค. ๊ทธ๋ž˜์„œ Px=0์ด ๋˜๋Š” x๋“ค์ด ๋œ๋‹ค. ์ด ๋ง์€ Null Space ๋ผ๋Š” ์†Œ๋ฆฌ.

๋‚˜ํ˜ผ์žํƒ๊ตฌ : det ๊ฐ’๊ณผ ๊ธฐํ•˜์ ์ธ ํ•ด์„ ๊ฐ„์˜ ๊ด€๊ณ„

A=\begin{bmatrix} a & b\\ c & d\end{bmatrix}
์˜ ์—ด๋ฒกํ„ฐ๋ฅผ \mathbf u=(a,c),\ \mathbf v=(b,d)๋ผ ํ•˜๋ฉด, ์ด ๋‘ ๋ฒกํ„ฐ๊ฐ€ ๋งŒ๋“œ๋Š” ํ‰ํ–‰์‚ฌ๋ณ€ํ˜•์˜ ๋„“์ด๋Š”
\text{area}(\mathbf u,\mathbf v)=|\,a d - b c\,| = |\det A|.
์™œ?
๋ฒกํ„ฐ ๋‚ด์  ๊ณต์‹







\mathbf{u}\cdot\mathbf{v} = |\mathbf{u}||\mathbf{v}|\cos\theta
โ†’ ์‹๋ณ€ํ˜•







(\sin\theta)^2 = 1 - \cos^2\theta
= 1 - \frac{(\mathbf{u}\cdot\mathbf{v})^2}{(|\mathbf{u}||\mathbf{v}|)^2}
u,v๋ฒกํ„ฐ ์‚ฌ์ด์˜ ๋ฉด์ ์˜ ์ œ๊ณฑ์€
A^2 = (|\mathbf{u}||\mathbf{v}|\sin\theta)^2
= (|\mathbf{u}||\mathbf{v}|)^2 (\sin\theta)^2
์œ„์—์„œ (\sin\theta)^2 ์‹์„ ๋Œ€์ž…ํ•˜๋ฉด:
A^2 = (|\mathbf{u}||\mathbf{v}|)^2
\Big(1 - \frac{(\mathbf{u}\cdot\mathbf{v})^2}{(|\mathbf{u}||\mathbf{v}|)^2}\Big)
๋ถ„๋ฐฐํ•˜๋ฉด
A^2 = (|\mathbf{u}||\mathbf{v}|)^2 - (\mathbf{u}\cdot\mathbf{v})^2
๋”ฐ๋ผ์„œ ์ด๊ฑธ ๊ณ„์‚ฐํ•˜๋ฉด








|\mathbf{u}|^2 = a^2 + c^2,\quad |\mathbf{v}|^2 = b^2 + d^2,\quad \mathbf{u}\cdot\mathbf{v} = ab + cd








A^2 = (a^2 + c^2)(b^2 + d^2) - (ab + cd)^2 = (ad-bc)^2
๋‹จ์œ„ ์ •์‚ฌ๊ฐํ˜•์˜ ๋„“์ด๋Š” 1์ธ๋ฐ, image ์ •์‚ฌ๊ฐํ˜•์˜ ๋„“์ด๊ฐ€ | det A | ์ด๋ฏ€๋กœ, ์ด๊ฒŒ ๋ฉด์ ์˜ ๋ฐฐ์œจ์ด๋‹ค.
n์ฐจ์›์œผ๋กœ ํ™•์žฅ๋˜์–ด๋„ det A๋Š” A์˜ column vector๋“ค์ด ๋งŒ๋“œ๋Š” ํ‰ํ–‰์œก๋ฉด์ฒด์˜ ๋ถ€ํ”ผ์ž„.
det A =0 โ‡’ A ์˜ ์—ด๋ฒกํ„ฐ๋“ค์€ ์„ ํ˜•์ข…์†์ด๋‹ค. โ‡’ ์ฐจ์›์ด ๋‚ฎ์•„์กŒ๋‹ค. (๋ถ€๋ถ„๊ณต๊ฐ„์œผ๋กœ collapse ๋˜์—ˆ๋‹ค)โ‡’ ๋ถ€ํ”ผ๊ฐ€ 0์ด๋‹ค.
๋ถ€ํ˜ธ๋Š” basis์˜ orientation์˜ ๋ณ€ํ™”๋ฅผ ๋‚˜ํƒ€๋ƒ„. det > 0 : orientation preserved , det < 0 : orientation reversed

Horizontal shear transformation with factor k

x์ถ•์€ ๊ณ ์ •. k๊ฐ€ ์–‘์ˆ˜๋ฉด ์˜ค๋ฅธ์ชฝ์œผ๋กœ ์ด๋™. ์Œ์ˆ˜๋ฉด ์™ผ์ชฝ์œผ๋กœ ์ด๋™.
det(A) = 1์ด๋‹ค. ํ‰ํ–‰์‚ฌ๋ณ€ํ˜•์˜ ๋†’์ด์™€ ๋ฐ‘๋ณ€์ด ๋™์ผํ•˜๊ธฐ์— ๊ธฐํ•˜์ ์œผ๋กœ ์œ ์ถ”๊ฐ€๋Šฅ.

Contractions and Expansions

๋ฐฉํ–ฅ๋ณ„ ์••์ถ• / ํ™•์žฅ

Composite Transformation

๋’ค์ชฝ ๋ณ€ํ™˜์ผ ์ˆ˜๋ก ์•ž์ชฝ์— ๋‚ด์ ์ด ๊ณฑํ•ด์ง.
์˜ˆ์‹œ. A์˜ ์—”ํŠธ๋ฆฌ๋ฅผ ๋ฐ”๊ฟ”๊ฐ€๋ฉฐ N์ž์˜ ๋ชจ์–‘ ๋ณ€ํ™”๊ด€์ฐฐ.
์ด๊ฑด ๊ฐ„๋‹จํ•œ ๋ฒ„์ ผ.

composite examples

y=(tan \theta )x ์— ๋Œ€ํ•œ Reflection์„ composite transformation์œผ๋กœ ํ‘œํ˜„ํ•˜๊ธฐ.
๊ทธ๋ž˜์„œ ์–ด๋–ป๊ฒŒ ํ–‰๋ ฌ์„ ๊ตฌ์„ฑํ•ด์•ผํ•˜๋Š”๊ฐ€? ํšŒ์ „ โ†’ ๋ฐ˜์‚ฌ โ†’ ์—ญํšŒ์ „ ์˜ ์ˆœ์„œ๋กœ ๊ตฌ์„ฑ๋œ ํ•ฉ์„ฑ ๋ณ€ํ™˜์ด์—์š”.
โ€ข
R_{-\theta} : ์ง์„  y = (\tan \theta)x ์„ x์ถ•์— ๊ฒน์น˜๋„๋ก ํšŒ์ „
โ€ข
T_x : x์ถ•์— ๋Œ€ํ•ด ๋ฐ˜์‚ฌ(reflection about x-axis)
โ€ข
R_{\theta} : ๋‹ค์‹œ ์›๋ž˜ ๊ฐ๋„๋กœ ๋˜๋Œ๋ฆฌ๋Š” ํšŒ์ „
๋”ฐ๋ผ์„œ A = R_{\theta} \, T_x \, R_{-\theta} ์ •๋ฆฌํ•˜๋ฉด







A =
\begin{bmatrix}
\cos 2\theta & \sin 2\theta \\
\sin 2\theta & -\cos 2\theta
\end{bmatrix}
์ž„์„ ์•Œ ์ˆ˜ ์žˆ์Œ.
๋ฐ˜์‹œ๊ณ„ 2D ํšŒ์ „์—์„œ ํšŒ์ „ํ–‰๋ ฌ์˜ ํ•ฉ์„ฑ์€ ๊ฐ๋„์˜ ํ•ฉ๊ณผ ๊ฐ™์•„ R_{ฮธ1}R_{ฮธ2}=R_{ฮธ1+ฮธ2}๊ฐ€ ๋œ๋‹ค.

Orthogonal Projection in \mathbb{R}^3

v = unit direction vector of line
Ax = proj_vx = (x \cdot v)v = v(x \cdot v) = v (v^Tx) = (vv^T)x

Reflection๊ณผ Projeciton ์‚ฌ์ด์˜ ๊ด€๊ณ„. ํ‰๋ฉด์˜ normal vector ๋ฅผ ์‚ฌ์šฉํ•˜์—ฌ.

ํ‰๋ฉด์˜ unit normal vector n์„ ์‚ฌ์šฉํ•˜์—ฌ, x์˜ n์— ๋Œ€ํ•œ proj_nx๋ฅผ ๊ตฌํ•˜๊ณ , ์ด ๋ฒกํ„ฐ๋กœ ํ‰๋ฉด์— ๋Œ€ํ•œ Orthogonal Projection Matrix๋ฅผ ๊ตฌํ•  ์ˆ˜ ์žˆ๋‹ค. ๊ทธ๋ฆฌ๊ณ  Reflection์€ proj_nx ๋งŒํผ์ด ํ•œ๋ฒˆ ๋” ๋น ์ง„๊ฑฐ๋‹ˆ๊นŒ ๊ทธ๊ฑฐ๋„ ํ‘œํ˜„๊ฐ€๋Šฅํ•˜๋‹ค.
์˜ˆ์‹œ1 : ํ‰๋ฉด x+2y-2z = 0 ์˜ Orthogonal projection matrix๋ฅผ ๊ตฌํ•˜์‹œ์˜ค.
ํ‰๋ฉด๋ฐฉ์ •์‹์˜ ๊ณ„์ˆ˜ = Normal vector (why? [x,y,z] \cdot [a,b,c] = 0 ์ด ์ˆ˜์ง์ž„์„ ๋‚˜ํƒ€๋ƒ„.)
unit normal vector =







\mathbf{u} = \frac{1}{3}
\begin{bmatrix}
1 \\ 2 \\ -2
\end{bmatrix}
Orthogonal projection matrix P = I - uu^T =
\frac{1}{9}
\begin{bmatrix}
8 & -2 & 2 \\
-2 & 5 & 4 \\
2 & 4 & 5
\end{bmatrix}


uu^T์‰ฝ๊ฒŒ ํ•˜๊ธฐ







uu^T =
\begin{bmatrix}
u_1^2 & u_1u_2 & u_1u_3 \\
u_2u_1 & u_2^2 & u_2u_3 \\
u_3u_1 & u_3u_2 & u_3^2
\end{bmatrix}
iํ–‰ j์—ด์ด๋ฉด, u_i x u_j ์‰ฝ๋‹ค ์‰ฌ์›Œ.
์˜ˆ์‹œ2 : ๋™์ผํ•œ Orthogonal projection matrix์˜ Rank, Nullspace, det
orthogonal projection์„ ์‹œํ‚ค๋ฉด ์ „๋ถ€ ํ‰๋ฉด์œผ๋กœ ์••์ถ•๋จ. ํ‰๋ฉด์ด image ๊ณต๊ฐ„์ž„.
rank (P) = dim(C(p)) = 2
Normal vector์—์„œ ํ‰๋ฉด์— Projection ์‹œํ‚ค๋ฉด ์ „๋ถ€ zero๋กœ ๊ฐ.
N(P) = span \{[1,2,-2]^T\}
orthogonal ํ‰๋ฉด ์œ„์—์„œ ๋‹ค์‹œ Projection ํ•ด๋„ ๊ทธ Point๊ฐ€ ์œ ์ง€๋จ. 3์ฐจ์› ๊ณต๊ฐ„์— ์žˆ๋Š” ์ ๋“ค์ด ํ‰๋ฉด์œผ๋กœ ๋–จ์–ด์ ธ์„œ Volume์ด ์—†์Œ.
P^n =P \to det(P)=0

conter clockwise \theta-rotation x์ถ• ๊ธฐ์ค€

์ถ•์— ์žˆ๋Š”๊ฑด ๊ทธ๋Œ€๋กœ ๊ฐ.
standard basis vector ์˜ image๋ฅผ column์œผ๋กœ ๊ฐ€์ ธ๋‹ค๊ฐ€ ๊ตฌ์„ฑํ•˜๋ฉด Rotation matrix ๊ฐ€ ๋œ๋‹ค.
1.
Orthonormal column๋“ค์ž„.
2.
Orthogonal Matrix๊ฐ€ ๋จ. โ‡’ Inverse = Transpose R_{\theta_x}^{-1} =R_{\theta_x}^T
vector norm ์ด ์œ ์ง€๋˜๊ธฐ์— ๊ฑฐ๋ฆฌ, ๋ฉด์  ๋ฒ”์œ„ ๋“ฑ๋“ฑ์ด ์œ ์ง€๋จ.
3.
์˜ค๋ฅธ์† ๋ฒ•์น™์„ ์ ์šฉํ•œ๊ฑฐ๊ธฐ์— det(R_{\theta_x}) = 1








R_x(\theta)

\begin{bmatrix}
1 & 0 & 0 \\
0 & \cos\theta & -\sin\theta \\
0 & \sin\theta & \cos\theta
\end{bmatrix}








R_y(\theta)

\begin{bmatrix}
\cos\theta & 0 & \sin\theta \\
0 & 1 & 0 \\
-\sin\theta & 0 & \cos\theta
\end{bmatrix}
zx ํ‰๋ฉด์—์„œ ํšŒ์ „์ด ์ผ์–ด๋‚˜๊ธฐ์— orientation์ด ๋ฐ˜๋Œ€์ž„.
(๊ทธ๋ ‡๋‹ค๊ณ  det = -1 ์€ ์•„๋‹˜. preserve ๋˜๋‹ˆ๊นŒ. -1์€ reflection์—์„œ๋งŒ)








R_z(\theta)

\begin{bmatrix}
\cos\theta & -\sin\theta & 0 \\
\sin\theta & \cos\theta & 0 \\
0 & 0 & 1
\end{bmatrix}
ex. 3x3์—์„œ y์ถ• ๊ธฐ์ค€์œผ๋กœ 30๋„ ํšŒ์ „
๋ณ„๋ง ์—†์œผ๋ฉด positive ๋ฐฉํ–ฅ ํšŒ์ „์ž„. postitive ๋ฐฉํ–ฅ์ด๋ผ๋Š”๊ฑด conter clock wise ๋ผ๋Š” ๊ฑฐ.

์ž„์˜์˜ axis of rotation

์Šคํ”ผ๋ฆฌ์ปฌ ์ฝ”๋””๋„ค์ดํŠธ๋Š” x์ถ•์—์„œ ๋Œ์•„๊ฐ€์•ผํ•จ.
์ด๊ฑด z์ถ•์œผ๋กœ \alpha๋งŒํผ ๋–จ์–ด์ ธ์žˆ๊ณ , y์ถ•์œผ๋กœ \beta๋งŒํผ ๋–จ์–ด์ ธ ์žˆ๋‹ค๊ณ  ํ‘œ์‹œํ•จ.
v๋ฒกํ„ฐ๋ฅผ z์ถ•์œผ๋กœ ๋งŒ๋“ค๊ฑฐ์ž„.
1.
z์ถ•์„ ์ค‘์‹ฌ์œผ๋กœ \beta๋งŒํผ ํšŒ์ „์‹œํ‚ค๊ณ ,
2.
x์ถ•์„ ์ค‘์‹ฌ์œผ๋กœ \alpha๋งŒํผ ํšŒ์ „์‹œํ‚ค๋ฉด z ์ถ•์ด๋จ.
3.
์—ฌ๊ธฐ์„œ \theta๋งŒํผ ํšŒ์ „์„ ์‹œํ‚จ ๋‹ค์Œ์—, ๋‹ค์‹œ Back์„ ํ•ด์ฃผ๋ฉด ๋จ.

\begin{aligned}
A &= R_{-\beta_z} R_{-\alpha_x} R_{\theta_z} R_{\alpha_x} R_{\beta_z} \\[6pt]
&= R_{\beta_z}^T R_{\alpha_x}^T R_{\theta_z} R_{\alpha_x} R_{\beta_z} \\[6pt]
&= (R_{\alpha_x} R_{\beta_z})^T R_{\theta_z} (R_{\alpha_x} R_{\beta_z})
\end{aligned}

Rodriguesโ€™ rotation formula

์ž„์˜์˜ ์ถ•์„ ๊ธฐ์ค€์œผ๋กœ ๊ณต๊ฐ„ Rotation ํ•˜๋Š” ๋‘๋ฒˆ์งธ ๋ฐฉ๋ฒ•. Standard basis vector ๋ง๊ณ , Orthogonal basis vector๋ฅผ ๋‹ค์‹œ๊ตฌํ•˜๊ธฐ.
ํšŒ์ „์ถ•์„ normal vector๋กœ ๊ฐ€์ง€๋Š” ํ‰๋ฉด์ด ์žˆ์„ ๊ฑฐ์ž„.
ํŒŒ๋ž€ ๋ฒกํ„ฐ : x์—์„œ (x๋ฅผ v์— projection ์‹œํ‚จ )๋ถ„ํ™์ƒ‰ ๋ฒกํ„ฐ๋ฅผ ๋นผ์คŒ. ํ‰๋ฉด ์œ„์— ์žˆ์Œ. v๋ฒกํ„ฐ์™€ ์ˆ˜์ง.
์ดˆ๋ก ๋ฒกํ„ฐ : ํŒŒ๋ž€๋ฒกํ„ฐ๋ฅผ v์ถ•์„ ๊ธฐ์ค€์œผ๋กœ 90๋„ ํšŒ์ „. (ํ‰๋ฉด ์œ„์—์„œ ํšŒ์ „์ด ์ผ์–ด๋‚จ.)
why? ์™ธ์ ์˜ ๊ธฐ๋ณธ์„ฑ์งˆ(๋ฐฉํ–ฅ : u \times v๋Š” u, v ๋ชจ๋‘์— ์ˆ˜์ง์ด๋‹ค. ํฌ๊ธฐ :








\|\mathbf v\times\mathbf u\| = \|\mathbf v\|\,\|\mathbf u\|\,\sin\theta
๊ทธ๋Ÿฐ๋ฐ ํŒŒ๋ž‘ v๊ฐ€ ์„œ๋กœ ์ˆ˜์ง์ด๋ผ sin\theta๊ฐ€ 1์ด๊ณ , v๋Š” unit vector์—ฌ์„œ ํŒŒ๋ž‘ ์˜ ํฌ๊ธฐ์™€ ์ดˆ๋ก์˜ ํฌ๊ธฐ๋Š” ๋™์ผํ•จ.)
๊ทธ๋Ÿฐ ๋‹ค์Œ์— ์ดˆ๋ก ํŒŒ๋ž‘์„ e1, e2๋กœ ์ƒ๊ฐํ•ด์„œ Orthonormal vector๋ฅผ ๊ตฌ์„ฑํ•˜๊ณ , 2D ํšŒ์ „ ๊ณต์‹์„ ์“ฐ๋ฉด
cos\theta(I-vv^T)x + sin \theta v \times(I-vv^T)x์ด ์‹์ด ๋‚˜์˜ด.
์ตœ์ข…์ ์œผ๋กœ ๊ฑฐ๊ธฐ์— ๋†’์ด ์„ฑ๋ถ„์ธ ๋ถ„ํ™์ƒ‰์„ ๋”ํ•ด์ฃผ๋ฉด ๋.
*v x (I-vv^T)๊ณ„์‚ฐํ•˜๋Š” ๋ฐฉ๋ฒ•
์ปฌ๋Ÿผ ๋ฐ”์ด ์ปฌ๋Ÿผ์œผ๋กœ ๊ณ„์‚ฐ ๊ฐ€๋Šฅ. ๊ทธ๋ฆฌ๊ณ  ๊ฐ ์ปฌ๋Ÿผ์€ e1-av ์ด๋Ÿฐ์‹์ธ๋ฐ ๊ฐ™์€ ๋ฐฉํ–ฅ์˜ cross product๋Š” 0๋˜์„œ av ์ชฝ์€ ์‚ฌ๋ผ์ง. ๊ฒฐ๊ตญ e1, e2, e3 ์™€ v์˜ Cross product ๊ฐ€ ๋จ.
*vector cross product ๊นŒ๋จน์Œโ€ฆ
determinant ์ฒ˜๋Ÿผ ๊ณ„์‚ฐํ•˜๊ธฐ.

\mathbf{a}\times\mathbf{b}
=
\begin{vmatrix}
\mathbf{i} & \mathbf{j} & \mathbf{k} \\
a_1 & a_2 & a_3 \\
b_1 & b_2 & b_3
\end{vmatrix}
=
\mathbf{i}(a_2b_3 - a_3b_2)
-\mathbf{j}(a_1b_3 - a_3b_1)
+\mathbf{k}(a_1b_2 - a_2b_1)

Translation

๋น„์„ ํ˜• ๋ณ€ํ™˜์ž„. ๊ทธ๋ž˜์„œ 2x2 ํ–‰๋ ฌ๋กœ ํ‘œ์‹œ๊ฐ€ ์•ˆ๋จ.
2D Translation์„ 3D homogeneous coordinate๋กœ ํ‘œํ˜„:

\begin{bmatrix}
x' \\
y' \\
1
\end{bmatrix}
=
\begin{bmatrix}
1 & 0 & h \\
0 & 1 & k \\
0 & 0 & 1
\end{bmatrix}
\begin{bmatrix}
x \\
y \\
1
\end{bmatrix}
์—ฌ๊ธฐ์„œ:
โ€ข
(x, y): ์›๋ž˜ ์ขŒํ‘œ
โ€ข
(x', y') = (x+h, y+k): ๋ณ€ํ™˜๋œ ์ขŒํ‘œ
โ€ข
h, k: x, y ๋ฐฉํ–ฅ์œผ๋กœ์˜ ์ด๋™๋Ÿ‰
โ€ข
์„ธ ๋ฒˆ์งธ ์„ฑ๋ถ„์€ ํ•ญ์ƒ 1๋กœ ์œ ์ง€๋จ (homogeneous coordinate)
์ด๋ ‡๊ฒŒ homogeneous coordinate๋ฅผ ์‚ฌ์šฉํ•˜๋ฉด ๋น„์„ ํ˜• ๋ณ€ํ™˜์ธ translation์„ ์„ ํ˜• ๋ณ€ํ™˜ ํ˜•ํƒœ์˜ ํ–‰๋ ฌ๊ณฑ์œผ๋กœ ํ‘œํ˜„ํ•  ์ˆ˜ ์žˆ์Œ.
n์ฐจ์› ๊ณต๊ฐ„์—์„œ์˜ ๋ณ€ํ™˜์„ (n+1)์ฐจ์› homogeneous coordinates๋กœ ํ‘œํ˜„ํ•˜๋ฉด:

\begin{bmatrix}
\mathbf{x}' \\
1
\end{bmatrix}
=
\begin{bmatrix}
A & \mathbf{t} \\
\mathbf{0}^T & 1
\end{bmatrix}
\begin{bmatrix}
\mathbf{x} \\
1
\end{bmatrix}
์—ฌ๊ธฐ์„œ:
โ€ข
A \in \mathbb{R}^{n \times n}: ์„ ํ˜• ๋ณ€ํ™˜ ํ–‰๋ ฌ (ํšŒ์ „, ์Šค์ผ€์ผ๋ง, ๋ฐ˜์‚ฌ ๋“ฑ)
โ€ข
\mathbf{t} \in \mathbb{R}^{n}: translation ๋ฒกํ„ฐ
โ€ข
\mathbf{0}^T: n์ฐจ์› zero row vector
โ€ข
\mathbf{x}, \mathbf{x}' \in \mathbb{R}^{n}: ์›๋ž˜ ์ขŒํ‘œ์™€ ๋ณ€ํ™˜๋œ ์ขŒํ‘œ

\begin{bmatrix}
\mathbf{x}+p \\
1
\end{bmatrix}
=
\begin{bmatrix}
I_2 & p \\
\mathbf{0} & 1
\end{bmatrix}
\begin{bmatrix}
\mathbf{x} \\
1
\end{bmatrix}
ํ•ฉ์„ฑํ• ๋•Œ
Made with Slashpage