You are on page 1of 15

Homework 3

Q1

Q2

Q3

Q4
Q5

Q6
Q 1

@)
Proof
suppose thae X is oprimal ,

let bean arbitrary direrrin


d
feasible ac x

xt od ε p θ sealar
,
:

posirwe

x oeL ly =
cxByassumprion
y
=
: +

c -
oid ' cx
y


=
Z

suppose thear none


of the
feusibledireurnwl at X Ep

reduces theoprimal cost. Then for


any ME P nehave


Y X (
y x)
=
+ -

Pefining o
:

diy -x and 1


y =
xtod

: ld ly clx
ciyzix
0 ≤ = -
:

b Rroof
Srppose thor X is
unique oprimal somrion toour

opimizarion prob ) m ,

Oxco'
franyotherytp
y

Using The same


argument
asin the
previons exerciseparr

→ cd =
cy - 20
c
Sinilary .

suppose thor
ony
otherfeusibledpeoriond arx

Increases the cost ,

If ne boneanother
oprimal someimY EP 4x
nTth
Cy =
,

then d
yX nich O nouldconsii-uele for afeusible
: =
=
1

direrrion which does not increuseThecost - conseradiction


a

Q 2

a)
Proof
Suppose thar
I is
empty , aned
supposefor thesake
of contraliorion
uwique
*
Solorion
thar x is not .there xists anoher
feusibll
dx* dy
* *
such thar
y -

Define afeusibledirertiond =

y
*
-
X
and
* consider the cost

'
intormsofthereducedCosL
change i d
TE ,
IEN

od =

iontiodi
:
di 20 for all iEN , andhortheremusr be atlnst 1

KEN such thar dk > 0

is
empryby exriiseassumpiion
I
:

,
reduced
very cost

ti
of iEN I =
N Is
posirive

did = tidi z ε
kdx 70

iEN

#
d contradiorionto dx = *

y
*

Proof
Suppose that 2 is an
opimal somriom
of The linnn
optimizarion

problun given in b) with an


oprimal cost ofi 2 = 0 ,

Suppose torthe sake of onradiovionthan


X
*
is nor
mique
s there xi5 tsanoher
feusiblesolwrion thar dx dy
*
yt such * =

Define a
feusible direution d * *
X T
y
=
-

,
. orrive ar a

Conradicriom nlutablish 2 faors


folling
,
,

I
isa fensible
*

y solurion for
.

⼩ )

WebaveAy =
and y z
by condivionsofthe
*
b

* 0
the

" ginal oprimizarionprob)


o∞
m

Represrtmg The
assumption ix * =
dytin terms of

he reluceL CostEE ,
IEN

cd = ε idi =
δidi = 0
,

iEN i ε N\ Z

wherltheslcond
theface thar
fuality tollons from
ti = 0
fori ε I ,

ti
'

:
20 Tor IE NI ,we obran di IEN
= o
for
and Tms
also Yi
*
* -
o f IEN J Cas xi =
0
fr

i ε N )
*
*
2
,
> x
Yi :

i εz i εz

Ne bave *
Y = asIEN ,
Y
, 0

Suppose for inesakl


for contradacrionthat YI = X ,

or
dr =
0
.
From dB =
+
B Aidi → dB = 0

IEI

: dulr = 0 x
*
*
a contradiorion
y
∴ -
=

thereexises a KEI such


har Y
, X aned
* *
ThmS

zYi ≥

i ε
zx .

The lasr lends the cortradicrion as we bve


assertion to

obrained xceeds assumed


objecriuevalne originally
an bear the

maximum of o .

Suppose Supposefor
X
*

unifre opimal sohution


thar is the .

somtson
z of
the sake
of coweraeliovion thar the oprimal
cisalsoa

fielols an
objrrive
vame of tzzi > 0
.

feusibll sowiim thl mal


'

Problem
orig
to .

From 尋
2i so there Is Or LUst one CEI Wi+n ZK 20 .

This faor immediately


rules ontThe possibilityfor τ= X

*
as X = 0
for itIEN .

Using ine eandard


5 cedure ser a
feasible dneorin e( =

pro ,

thefeasibie
* *
onel consider clddlong
y -
X theCo3r
change
direcriond m

terms
ofthrreduced CO 5 LEi ,
EEN
tieli cadi
-

'
d
-

= > 0
c =

iGN I ε NI

where faor ti
seond fuality follows
= o
The fron the tnar

for itI and the lasr inequaliry follors from


the fact that

*TsThe
X

uniqueoprimal
sohurion .
From this
Tneqwality ,
ne

IEIN should nonzeroli 7


see tharar Jeusr one di , I be

i
for some
:

X thar20
* Tor implies
70
=
0
I tN
,
this

IEN I .
Br thi, is a conradi otion to the facrthr

2 :s
fusimle for ) ,
as it Violares the
Secondquality
constraint ,

3
,

@
min - 2X
1
-
X2

St X. -
X. + X, = 2

X2 X
64
=
x, + +

x
X2 X 3
.
, , .
xg

min
00 - 3 x

⼀创
S t

]
-

Ti
.

! !
Xs

x
4

x= J
00
, 2
,
01 is a basic fensibbe solwrion
accordingly
lor

B⼩ =
3 and ( B ( ) 2 =

4 The -

corresponding basismatrixB rs I
↓ initial tablenw ;

X ,Xa3 x
4
0 2
— 0 0
-

*
X; = 2 l 时 1 0

× 6 1 ( “
4
=

X Xa Xs X
, 4

4 0 -
3 2 0

x = 2 1 + 10
1

*
1
×
4
=
4 0 2
+

x Xc Xs X
( 4
| 3
( 0 0 0
2 2

1
1

(
X
4
= 0 2
1 2

1
|

X 2
2
= 0 1 -

2 2


The basio feasible solnrion is 0
2000
= 4
,

X + X2 ≤ 6
1

x
*)
) 21
4 , ,
0
,
0

D
X
x

1 (
(

0 2

X -

Xc ≤ 2
.
4
.
Phase I
min Xor XrTX 8

st x . + 3 ×2 X0 2
τ 4 ×4 + X5 +
=

x, + 2 X2 3× + X
4 2
-

5 +
XT
=

-
x
1
-

千×2 + 3×
3 + X8 = 1

x1 X2 Xs X x5 X0 X7 X
,
, ,
4 , , , ,
8 z 0

x, x2 xs × X5 X6 X7 x
4 8

- - 1 -
- 3 - — 0 0 0

x
0=
2 1 3 0
4 1 ( 0

l
xy
2
=
1 2 0 -

3 0
c

×8 =
1 千
*

- 1
0 0
3
- 0 0
1

7
π x2 xs X
4 45 X 6X x


-
2
- 5 0 时 2 0 0
1

x
0 = 2 | 3 0
⼿ “ 0

l
xy
2
=
l 2 0
— 0
1 0

|
X =
|
4 1

33
3 | 0 0

3 0 0
3
2 enter Xy
X

xi 5L

¢x x2 x3 X
4
x5 X6 X 7x
0 0 1
T 0 | 0 T 2 0

x
21
= | 3 0
4 l ( 0 0

= 时 1 0
xT 0 0
⼀ 0 +
1

4 | 「
1
X3
(
=
0
⼀ l 3 0
了 3 了

x x2 xs x x5 X0 X 7
4 x

0 0 0 0 ( |
0 0 1

xF 2 1 0 0 ⼀

( | - 2 3

X2 = 0 1 0 9 0
( -
1 0

1
1
1 2
1

×
13 -

0 0 l 3 3 了

3 S

Phase 丘
v

x. X3 X5


X
1 4

0 0 0 3 -
5
*
1
0 0 - (
T 1

x
2 = 0 l 0 T
1
"
x =
⺦ 0 0 1 s 了
3
x, xa X3 X
4 X5

3 5 0 -
82 O

X 1 O 1
7
= 2 0 、
5

*
x
2 = 0 1 0
T 0

1 28
x = 01 1
33
-

3 了 0

π xx X3 X
4 X5

8
3 5 0 0 0
τ
1
x
25 =
1 0 0
|
T
1
x = 0 0 1 0
4 T
1 |
x =
、 T 1 0 0
33
3 -

Wehone reachell 5
*
non the
pome X =
( 0
,
0 ,
,
0
,
2 )

The
opimal Cost is 3 ,
applied
5.Phase I of smpex meshod
canbe

) a using initial tableww ,

this as an

represent feasible
:

Thus tablean muse


a

Somiron musr be
,
Tts neries
nonnegarive
→β
z 0

The arrificsal variables hore been driven oc

and therearlno
further restrTorions ,

b , Thefirsrrow in
the presurttab )un Tndicats
thar the
problom is
infeasible ,

Ne must Tndicatelck
he
firsr ron can
infeasibility ,

0 0 0 0
f 3 r E

2 1 3
。 1 0 0
0

X0 α X
β= 3 XT
+ + X5 +
4

X
an
Xa= 5 7
they
s
( 4 .
o
as are
, X
X

non _ basi

holods bo moirterwhich valne


? The
equarion
and
β
2 assume ,
o) The
correspondingbasic sohotion is
feasible bot ,

do have basis
we nor an
oprimal ,

: The tebleun
represenr fensible lwrion
'

musr a
,
its

unevies be
nounegarive
musl

β
20

: Thesolurion be able reoluce


'

isnot to
oprimal ,
ne musr

CoSC

arlensr
one of thecondirions So , ro r Eco

ds toblean must
' :The
represerr feusible lurion a
,
its

eruries be
monneyarive
musl

β
20

the
ofthe
:

oprimal coslis - D
, no componenr column

enteringthe next
step con be
possirive ,

Upon closer

Tuspection onlycanlidare
ne see thar
,

is X anel the
4 ,

conlirion sarisfierl
is
by serring X ≤ 0

Tomakl surl tharthe5


mplex dlqor 7 thm chooses x
4
to enter in the nexe
step ,
ser 800 ,
r 20 ,
ε 70
,

: The toblean
e' musr
represnmr fensible lwrvon a
,
its

3 be
nonneyarive
musr

β
20

For be candidare
X6 to a
for sbe basis ie
wrering ,

reeluced be
negarive
cost muse
V < o

Fos Xs to exit the basiswhin Xs tbe zar


:0
inturs ,

tesr muslrule oor in favourofthesecond row

个 J , =
*
→ O = min
B }

R n ar 3
4
f

) For Xo to be a candidonefs
iwrering
the basis ,
TIs

reduced musrbe rive


cosr
neyu
- Ec

valne remaing
ske
only way
thar the overall
objecrive
unchangel is when X, Uxitsthe basiswithonr
changing
the cost wherweadd the
firsr ron to thl zeroth

ron wi+ h
peramerer
β=
the
0

Q6
.

0 .
… Emei … . Cn

… 0 …
x
acime (
)

1 a )
1

:
:
: …

: : :
1 .

Xm acm
t - am
o
n)

, +1 ) .
,
m

@ 已知 τ 20 V (m 1 )
j ] (
=
1 )
Enco
-
,
… n -

,
且只有

minimirarion
problem

只有 Enco ,
其他皆非負
∴ Ci 為追入 basis 的唯⼀候選⼈
且此為 , ron degrnerare
θ
→ * t

Xn inrers basis θ *
→Xn + * ( =
Xi ,
θ t 0

Xnincreases co an vame s 0
Xn ' > 0

'

Jer Z > Z
basis anel ' ivor is
by ; Xn mnsr rover
p degnerare
*
θ =

xn
*
*
= Xn +θ ( ,
θ = 0

degmerare
' ,
sohirion
Tim X n
=
Xn

E = E '

basic
Xndo nor hoveto be the
only one
oprimal
feusible solurion .

You might also like