Four Subspaces of A matrix @ xn )

( T C CA ) Glum space → ERM

2. → CC AT ) Rsu
Spee - C- Rn
I
g
-
N CA ) Nullspeee A*= - →
C- Rn

left Erm
CAT ) Nullspeees
ATTY
.

4-3 N -

⑤ Am I
m

47¥
't R

✓ ! row

re
here
direr
y
-
r .

.
.

e-
aa.E.hr
Fudanertol Those of lines Ayden
PART I .
 -

ORTHOGONALITY -

88=0
-

ET orthogonal
Te
.
- -

ai ET -43 . -

ate .

adf.bg;)
Can -

-
aiataibetasbs .

a. b=aTb .

ORTHOGONAL VECTORS ? VTTU =D → N D

,

are
orthogonal vectors
.

ortwgol Vector speed !

NEF
og
→ o
c.CA#) - Row spear .
A-
l
Roa=o
o

Rou2 I 0
Definition's :
Tun subspaces D and DM of
a vector space ee
orthogonal if every
week

of D is perpendicular to the
space
every
u

vector ~ from V Spee -

NTD = UTV = O for all w E V

U E D .

2%1%1
→

Ffg I subspace .

T which vetoes )

€#he nothin EFF Tahoe ?

'
n'
-

[of]
and line
Xy plane t
u = E xym
-

V
one
portugal speech .

=
EE) - E z - err UTU = o .
 ÷:÷÷÷÷
Corney
.

.TN#iF-aTaiqn-rr: : E:Iym-En-r
1543--0
Ahllspaaisortwgorrwpe =
ng
 Left nullspeee
AMY a- o

) IE.LI?.e.AnifLyf.o
* *
.

di - = r

\ NCA )
column space is
orthogonal YE
to the left nullspaee .

darn = m - r

ORTHOGONAL GMP LEMEN T E

-

Definition The
orthogonal plant of
.

as - a
.

D vector the t is
subspace contains
every
perpendicular to subspace Tl . V is
*
orthogonal as -
planet of D , HI$& V , =D
¥E#÷ . . * .
.

i÷÷÷÷÷÷¥÷÷÷
 Fundamental Thes- sf
Pert I
Lmao
Algebn
-

complect of
- NCA )
row
is the

space
orthogonal
CC AT) in Rh
the

esurient of the
-
NCAT ) is the o-
thyroid
Colin speee CCA ) in Rn '

-
-

NCA )
A x=$ -
# o

@XptXspLtrwtspeciel.A
* =

=b

Acxptixg ) z
Axp + Axs = b .

y
-
 )
fiats :
*

inverts
N t - invertible

Velu Deco pother pseudo inverse

singular
-
-

if A is wt intertitle → find are to be

invertible its inverse Ihor Reverse of
→
assign
A

Combing Doses fro - subspaces .

X - Xr* Xn Cxzxptxs )
£ &
nulls pace base .

rows per
nose .
CC AT I NCAA )
 How wear we mite vector & as perpetual
• carbine then of subspace 1
N
"
Hot
tilt
V?
not proper
 PROTECTION S .

ctioitontu the

) ¥2?
' - x axis
b is

Pi =

[I }
Pz :projection of b on h th

yZ pl
- -

PE
[§)
I
i.iii.
.

Htt'd
Ps .
 put;) -

to.
I
.

Petre =

TI ft (g) = b

(
.

:S
Tk k Pz .

Pdt Pz = If

¥¥÷
Ps b t P
 bin
Projection onto a .

P
in
-
"
e-
*

¥¥÷÷ .

1
PILE
a. CC -
Ia )=o .

P:
projection E=a
To
Etta 2
.

p=a
.

a
?
.

P= p . e=o
ate
F : scalar number
(b Foa ) o
-

p
-

projective
=
.
.

a. (b -
Iola -

÷:-. *=¥ ¥I
 r-og.is?s=TpTIp:=aa.:tIf
matrix
Projection
.

Exempt .

.
Who 't is the
projectile of

=[§) on the
the
2- axis

prsyction
ad
matrix
whet is
fo - z - axis's
-
boris rectum

a=[§)
for -

E -
exit
Subspace t axis

Penna at @ Koko

.at?a.=Yo:!?e*3)-= ?
otcb
fl
=,
-
=

e -
-
b -

p
-
-
b - xn . * =

=3 -
psi
f:p
. - s -

:÷s
tooo ]
' =

p='
to
RP
fog ! ? )pnza
-

Pa -

fog)
 b
J}) ufhiind pmyeetio
-

- its onto

a line
?
-
-

¥'÷→

E-IE.ie#=.t=zP- 2L7J-EI3e- b-p- f'jf- I - E

.

?]
ti
esp =
? e-p -

-
on
[ =L
 Projecting onto the Subgpee -

b .
what is the
projection of

µe=b
--µk.¥
P
-

b on the space defined

'
III. 3-
is nano -
no .
9
.

In
.
.

an
•
2 .
A
e OE on
-

) .

onto the alum

projection
of
ou At
?
ont lore
'

Penna c- for .

p - AI = a. ni torta . . oniin I

e -
- b -

p e is perpendicular dohn spee off .

 Each basis vectors of CCA ) must be

perpendicular to the error week .

aids - Axa )=o

A-
Eyal - -

on ]
Azt (b AIL
A'
- - o .

int Cb -
AI ) = u .
-

P ACATAJIAT
It
#
(b -
Axn ) =D
projection
vets
A- TAI
.
.

A- Tb -
= o -

Normal Equation
*=anIb '
Atb ATTA
b

CATA )
.
-

'
A CATA ) Atb
a- -

= -

#
Steps
.

'

1- First I I # TAJ Atb

ly =
'

2-1 Calculate pmyefifionoecp = AT

7) Determine projection netrin P = ALATA 'TATT

p =P b .

initiate ?
fit ]
*

E : :3 t.EE :3
,

=
Projection ortho CCA )
=p, =P , b .

projectiles onto error

plane .

Pz =
Pzb .

is

:* ;
.

me

- e
belongs
which spae?
I1$ ( / Left mllspee is
.se#ehgm
( CA ) to CCA )

ee CN EAT )
P E CCA )
 onto Row of
projector space A

Pr =

E=b Pr → Error
beliefs togging?
-

ere e E Null space

e e C- NCA )

Pre CCAT )
Row sp -

X =
Xp +
XS

rotsp
.

- xfnmiksp -