Lecture14 L0203 Marked
Lecture14 L0203 Marked
1. Dimension
2. Rank
[0 1 ∣ 0]
Indices of free columns
1 0 ∣0
Insert (n-r)=2 rows to
decide which rows to insert
Free column:
[I | 0] =
become row 3 and 5
column 3 & 5
AX-bffAX-0.no
501 →
Thereisanexamplethotthedumnpaeofacertahmatnxca.be
snifed.tl 渄 啊
RetThndoesNOTme.cn
clmlonhsiplifedsinceMmgharepeeidstmcture.weneedext.ro
studiednext 。
1st be
Review of Linear Independence and Basis
n
Eg1 {e1, …, en} is a basis of ℝ . (called “standard basis”)
Eg2 {Eij, i = 1,…, m; j = 1,…, n} is a basis of ℝm×n.
Part I Dimension
Strang’s book Sec. 3.4
NextR3.eg.la
InterestngQuestin7.Thereremfnitelymcybose.fr
ez.es/i ab si ,1ei, 2,eit yrsab n,lei, .tez,e.t zteyisab sn.Questim_e.
,
QIWhatsthebiggestbas.37Vot.fr
chosejgchoosecivotef.cl
Qli 100%
% choose % doe
" "
" "
2 :
8 。 ,
20 3 ,
Size of Bases ⽐ ⻔
上.my 们 们 们。
+ +
srnunv.A
Recall:
1
0
ℝ T
1
0
1
{[0], [1], [1] } is a basis of
a-A-o-n-t.v.ie
-1 :)
1
3
.
到 ⼆ 0
Ay
,
a taetas-authlbtdtslyEYln .nl
=
1 -l m e c r l y m de p e nde nt i ui l s h oui m p os i bk .l t ? S h ow Ii
Claim: The set of four vectors {v1, v2, v3, v4} ⊆ ℝ3 is not a
basis of ℝ3.
-01 但
-
3
11 :
2
) 问 沿⻔
arehneagdependent.li
, ,
舆 showlivi.IE/R2arelmeerlydpendent
solutiontExerciselSMore-S.SN
[ 溺 ⻔ , , [ 当张
lmedgdepende.at ,
iii) + ⼈
问 ⼗ 多
闹⼆
0ha
non-zerosolul.tn ,
Axwhasnon-zerosolutim.lv
(iii)
满⻔ - 16
年间 -
1 6 - 孔阔
XF-MX3.la
X,
1
+
1.x 0
'
⼆
piw
'
仙 t
2 "
P '
3
Solut.int Exerc.se 2
cdumn.at
3 ( n= 3)
A ※
1
mostzpiwtslrEmzjfrowopennii.IT
:
-1国
2 rows.3columns-atleastonefeuoni.ba
x-mcnysolutions-columnsofzsmatrixarelme.org
dependent 。
Ge1iemnaet嗨
rn.cxal/r2elemseual.tlyG/7luEmPENDENT!
CommonPathtDeadEndAfewstudentstriedtgenerd.ie
thepooffExerciseltExercisezosfl owsilb.giab.ci
bthienbkblza-q.ci/Coansn-Butiderdifer tnhtceents .Ihfbggeti z=O,tshetnublcksmaiobk;IfCz-fGt=stoconpl.eu
acbz.cz 0
0
,
ImportantthngiTthespecificnumbers
theSHAPElmcnithusranmalters.NO ,
Key Lemma
Lemma 14.1
Suppose m > n . Exercisezfnomtntoayn.lemmol4.lgeneral.us
Grdlggeneral.us
Corolkytanylmeorsp.ee 。
vii.
u1, …, um are linearly dependent.
5th
notwntein nasamatnx.smcetimgnotbeave.ci
丛
n
Corollary: Any (n + 1) ℝ
vectors in are linearly dependent.
or
疑
然 !0min
,
enis-D.it
剾 ,
→
Zmarelmeerlgdependent.lIZ.zziizmJENnisanidemotridi.li
SousetheresultfR.la ⼀
,
,
X.s.tsxntntxnZFO.VerifgyourselfiXU.tl/zUzt-tXnUn=O.
⺕
Exampleoflor espondingleiRhosabasntu.us
, v41 :
[ % 了 1 % 了 [% 了
, , ,
诎了 ,
以 =
[f i ] = V1.2 t V2 -3 tls.lt V40
Thenzi.fi/isthecorrespondngUctiunderbas.3lVnn,
1 6 ,
V3 V4
, ) 。
Z ,
Dimension = Size of Bases
4 (bases have same size)
Theorem 13.1 First “big”
If {u1, …, um} and {v1, …, vn} are bases of a linear Theorem
space V, then m = n .
s
in this course!
prwfofThmM.ee
lusiglemmal4.DAssumemm.u.in
,
…
,
umarelmearlydependentlbylemmal4.in ,
NOTabasis.Contadictim.im
ns.miky.nem.Y-m-n
≤
。
Dimension = Size of Bases
Theorem 14.1 (bases have same size) First “big”
If {u1, …, um} and {v1, …, vn} are bases of a linear Theorem
space V, then m = n . -
in this course!
n-drmensiondspaelRnweusedthstermjbefrel .pro
Remark: If there is no finite set of elements that can span V, we say V
-
阎犫
# "
亚 0th
肶选
lme ?
# ofin 裨
fpomnhe.i-lmcng.cat
9ᵗʰ
"
)
。
Dimension Measures “Size” of Space
93871 ⾏ !刴
→
Option 2: Dimension.
Fundamentaleerspaaperpec.tn
difenencebetweenRandmllfoml.in
e)
segn-3.io
Exercise
tshesshstsststsehbhstosstrtfotohbhessFormabo. is
. 。 ,
triangular matrices
lsponlflm.ihdg.li.dinlVF3.lk
Space of n × n upper triangular matrices
1 % 了 [% ) 后 9 7 50 d.mu ) 3
Basfr 2×2 =
: .
.
, ,
6
1 31 8
的
1 3 3 ×
I 㸚 :
了 ,
dm =
.
T 以"
iEijijgfrmabasis.si?e0fba.
:
了
k-inlntn.dnlpnl-ntl.M.e-n-duklRlki.se
些 2.x ⽔ , ,
⼀
, 划 .
着
1 0 a 0 c
0 1 b 0 d
C
0 0 0 1 e
0 0 0 0 0
nextp.ge ,
SolutionloExerc.ie (5)
1: ⽐ ! ! 1
"
…
'
0 0 0 i e
t.VE V3 V4 V5
lasttnmeiArecolumnsdependentiYeoobserualim.la
jdtb
) ( b) 竺 红 0
-
⼆ V2 巧
1 (c)
.
ntkdmtteluty-T.lutdvzteq-gily-aeib.iyce.ua did.ie
qanlnnnn_i-t-g-spania.TT 㒣 ) 呼呼 :
"
Quicker Way to Check Basis
Proposition 14.1
Consider a linear space V with dim(V) = n.
1) Any linearly independent set can be extended to a basis, by adding more
linearly independent elements if necessary.
2) Any spanning set in V can be reduced to a basis, by deleting some elements
if necessary.
⼀、
1
[1]
Example. v1 = 2 is not a basis of ℝ3 .
Weexpanditabasis.StepliFmdaithatc.cn
NOTbeexpressedbyti-checkfn.ij-siopdifnotbasis.cn
tinue
thotenNOTbeexpressedbyoii.ch
Etepnntndaū
akfi.i-l-sopfnotbosinontinue.no
Whenstoppedwegetaba.is ,
Expans.in?Theproces ofexpcndnganlm dep.set abasi kilEV
HOWMcngg.sn
↓
Whot.im?
亿州 ≤ V
ㄓ
mmustben.hu
↓
问 ,
V2 , EV
!
亿 , v2,
…
,
则 basisof V.
Processofshrinkingfnningsetspanningsetfu.tn
vmievfhechidgendentiy-ASTopwhensiop.wegetamin.mu
⼀
,
,
matonedement.tt
spcnningsettvii-u.IE V.
IV.
lili .im/EVIt'sabasis.fhech:dgendent?
⼀
,
,
kn.y-ASTopcutoneel.me
幽
nt
i
Quicker Way to Check Basis
Proposition 14.2
Consider a linear space V with dim(V) = n.
If u1, …, un are linearly independent, then they form a basis of V.
If u1, …, un ∈ V can span V, then they form a basis of V.
Remark: For n vectors, only need to check either “span” or “lin. indep.”, not both.
1 2 2
[0] [0] [1]
E.g. 0 , 1 , 3 are linearly independent, thus a basis of ℝ3 .
"
noneedlodd 垒
ang3lmecrlgndgendentuectnrsfrmaba.is
了
In R ,
。
Quicker Way to Check Basis
Proposition 14.2
Consider a linear space V with dim(V) = n.
If u1, …, un are linearly independent, then they form a basis of V.
l gc omb.i e
If u1, …, un ∈ V can span V, then they form a basis of V.
Remark: For n vectors, only need to check either “span” or “lin. indep.”, not both.
1 2 2
[0] [0] [1]
E.g. 0 , 1 , 3 are linearly independent, thus a basis of ℝ3 . ,
getwhat7.nvectrs-rl.es/tx=ohanonon-2ero-sohnn
Recall: Lec 13 Part II:
n × n matrix A is invertible iff the n columns are linearly dependent.
From Fact 14.1 and 14.2, this is equivalent to:
a) Its columns form a basis. x A 0 0 ntheuniquesoh.tt
=
-
1 =
b) Its columns span the whole space, i.e., Ax=b is solvable for any b.
眑
等vec.mg
basisnm_e.thuert.be
nvecmdg
Invertibility Conditions Only
conditmshuolvenumberil-u.no
2
) 4 2 1 4 4 .4 ( 1
Let A ∈ ℝn×n.
The following statements are equivalent:
1. A is inver>ble
2. The linear system Ax = 0 has a unique solu>on x = 0 Equation solving
3. A is a product of elementary matrices
4. A has n pivots; or equivalently: rank(A) = n Rank
-5.1507nsec
7. The columns of A form a basis; basi.tn columnsy Basis
ii.
8. dim(C(A)) = n..
Dimension
npiuots-EAx-ohosonlyIsolutiono-ncolumnsofAarehnea.ly
rank(A) = n <==>
mdepende.at
ncolumnsofAformabas.is
dmlc (A)) = n .
Relation of Dimension and Rank
⼀
Anything interesting to guess?
Relation of Dimension and Rank
ˋ
井
Analysis: Need to consider the case of no unique solution.
-
" "
⟹ Ax=0 has non-zero solutions. 甇;嚟 第
笔
"
,
.net
cntioltrol.hn
thepnof 。
-
Action: So, let’s revisit the non-zero solutions of Ax=0.
ForRREFi.klRkdimkcmlpiuotcdumns-OC.to
1 oooool.tn#0fpiwtcolamns
NOTC01hcidenal.it
3 315 ⼆
eg q ronhlmd.hn KCRD
R =
0 0 0 1 e
# 尛 " 1点 "' "1
,
少 。
whataboutgeneralRREF7.cn
IF.li !
!!
"
※
lf-tgc-sponlta.nl )
piootcolumsfreecolumnsolumnsfFtspoultee-erjl.CI
TrgenerdRREF.thepi¢ votcol
) ,
thusd.hn
) umnsfrmabsofc.
ofpiuotolumn IR
(R = # ,
= # ofpiootstnkk) ,
Revisiting Solution Set of Ax=0
Pivot columns Pivot variables
x1
1 0 a 0 c x2 0
Linear system: 0 1 b 0 d x3 = 0 .
0 0 0 1 e x4 0
0 0 0 0 0 x5 0
Free variables
Free columns
The complete solution is
x1 −a −c
x2 −b −d
x3 = α 1 + β 0 , ∀ α, β ∈ ℝ .
x4 0 −e
x5 0 1
Plugging in solutions: −c
−a 1 0 a 0 c 0
1 0 a 0 c −b 0 −d
0 1 b 0 d 0
0 1 b 0 d 1 = 0 0 0 0 1 e
0 =
0
0 0 0 1 e 0 0 −e
0 0 0 0 0 0 0 0 0 0 0 1 0
0
ns.t so-s-stswilprovelemmel4.znext
Fact 1: In RREF, free columns can be expressed as LC of pivot columns.
。
Elementary Row Operation does Not Change LC Relation
Pivot columns
Consider the matrix R̂ that becomes R after one step
−a
1 0 a 0 c −b 0 of elementary row operation.
0 1 b 0 d 1 = 0 1 0 a −β c−βe 1 0 a 0 c
0 0 0 1 e 0 0 βR3+R1 0 1 b 0 d
0 0 0 0 0 0 R̂ = 0 1 b 0 d ⟶
0 0 0 0 1 e 0 0 0 1 e
0 0 0 0 0 0 0 0 0 0
Free columns
Claim 14.1: Elementary row operation does not change “linear combination” relation.
i.e. if α1 ⋅ (Column 1) + … + αn ⋅ (Column n) = 0 holds for M,
then it holds for the row-transformed matrix too.
emailiii)
-
M
Rigorous proof? u.cn ] [
7 xmtn_n-gthx.int i
.lt/t.
邀
t.IM 的
䪉
⼆
EM.fm
M.at?thenx.(col1ofNt--+xnldnofaiFO.
IxnkdifMt-txn-lalr_nofmw.net
Egf
ni.a-OM-EltEO-O.lt
哟
ˋ
keyideaofPnoflinecr ombinationofcolumnslghtmuy i.NO
iafectedbysrowoperationsclefmult.pl g)
FinishhgProofofdmlclAD-r.Useaa.hn/4.l(r0woperaNndoaN0T dongeLClhneaanb.nax.in ) ,
lopwved.nl d.hn
KCRD-r.lt
(A) ) =
Ri-HHPH-HHHIRREFYifreeolu.ms !路
fponfre ceoeollumurnnsnsofpivoRtcofdAufmornmsoabasifA_psiwoftCcoMpilunomtcoduonsof
fRrpiotctl..tins
p
pafeecolumnsof A
mdependentEPNtchnsofRarel.n.mdependentidmlc.CAT
r
h
Definition: Row Space
Definition 14.2 (row space)
a⊤(1)
Suppose A = … ∈ ℝm×n . is a linear space.
a⊤(m)
Then span({a(1), …, a(m)}) ⊆ ℝn is called the row space of A,
r
denoted as Row(A).
sfg
Eg: C(In) = _________
R
1 0
[2 3]
Eg: Row space of A = 4 3 is the set
⼀
三
{α1 [1,0] + α2 [4,3] + α3 [2,3] ∣ α1, α2 ∈ ℝ}.
⊤ ⊤ ⊤
侧 们 《煳 ,
What is dim(Row(R))?
Pivot rows
Non-pivot rows
灃11 0 a 0 c
0 1 b 0 d
0 0 0 1 e
0 0 0 0 0
=
u⊤3
01×4
.
u⊤1
u⊤2
Why?
。加
Together: dim(Row(R)) = # of pivots = r.
Elementary Row Operations do Not Change Row Space
At-R.si?u,u)=spanlu,uy-Oj=-
Consider the matrix R̂ that becomes R after one step of elementary row operation.
dim(Row(A)) = r.
dim(Row(R)) = # of pivots = r
Rank Theorem
o-RISomebo ksusedmlcADordimlRowlADasthedefhitm.fi
Theorem 14.3
Row-rank, column-rank and rank of a matrix are the same, i.e.,
rR(A) = rC(A) = rank(A).
Equivalent expression: dim(Row(A)) = dim(C(A)) = # of pivots.
iink
"
,
Example
器
rR(A) = 3.
C(A)= 。
R=
rC(A) = 3.
3 pivots,
RetclAHCCM.checkyourself.Butdhk.CA
So rank(A) = 3.
-
Summary Today (write Your Own)
Detailed summary:
Summary Today (of Instructor)
ottpeoeososhp.stizeestisrt s h.ch
Dimension, Rank and Nullity.
,
Detailed summary:
1. Dimension
—Dimension of linear space = size of any basis.
isthefundotnnofdefnithfdnens.tn
—Theorem 14.1: Any basis has the same dimension.
"
) ,
2. Rank
Column/row-rank = dim of column/row space
Rank theorem: row-rank = column-rank = rank (= # of pivots)
Bstssstttssssststttts
3. Nullity
—Nullity = dimension of null space.
—Nullity theorem: Nullity + rank = n.