Upload
osborn-fletcher
View
218
Download
0
Embed Size (px)
DESCRIPTION
The trend effect that Kate, Duchess of Cambridge has on others, from cosmetic surgery for brides, to sales of coral-colored jeans.” “Kate Middleton effect Kate Middleton effect 3
Citation preview
Lecture 2-1
Submodular Maximization
Weili Wu Ding-Zhu DuUniversity of Texas at Dallas
Outline• Kate Middleton effect• Max Submodular Function
2
The trend effect that Kate, Duchess of Cambridge has on others, from cosmetic surgery for brides, to sales of coral-colored jeans.”
“Kate Middleton effect
Kate Middleton effect
3
According to Newsweek, "The Kate Effect may be worth £1 billion to the UK fashion industry."
Tony DiMasso, L. K. Bennett’s US president, stated in 2012, "...when she does wear something, it always seems to go on a waiting list."
Hike in Sales of Special Products
4
• Influential persons often have many friends.
• Kate is one of the persons that have many friends in this social network.
For more Kates, it’s not as easy as you might think!
How to Find Kate?
5
•Given a digraph and k>0,
•Find k seeds (Kates) to maximize the number of influenced persons (in one step).
Domination Maximization
6
7
. toequal isunion whosesubsets theof find,0integer and},,...,{set ground a of
,..., subsets of collection aGiven :Cover-SetMax Domination Cover-Set
hard.- isMax Domination
1
1
UkkuuU
SS
NP
n
m
pm
Theorem
Proof
1S 2S mS
1u nu2u
ji Su
nodes influence seeds solution hasCover -Set knk
Outline• Kate Middleton effect• Submodular Maximization
8
Max Coverage
Given a collection C of subsets of a set E, find a subcollection C’ of C, with |C’|<k, to maximize the number of elements covered by C’ .
Influence Maximization is a special case of Max Coverage.
Max Coverage
Given a collection C of subsets of a set E, find a subcollection C’ of C, with |C’|<k, to maximize the number of elements covered by C’ .
.||)( define , of ion subcollectany For
SAfCA
As
Submadular Function Max
11
kSVS
Sf
Rf V
|| subject to )( max
Considerfunction. submodular increasing
monotone nonegative a be 2:Let
Greedy Algorithm
12
.output };{
set and )( maximize to choose do to1for
;
1
1
0
k
ii
iv
SvSS
SfVvki
S
Performance Ratio
13
solutionm. optimalan is * where*)()1()( 1
SSfeSf k
Theorem (Nemhauser et al. 1978)
Proof
Proof
14
*).()1()( i.e.,*),( Hence,
1 )11(
).(Then ).(*)( Denote))()(()(
)()()()(
}),...,{(
}){()()( *)(*)(
}.,...,,{* optimal Suppose
110
1
/11
1
1
11
1
21
21
21
SfeSfSfeaea
exaeak
a
aakaSfSfaSfSfkSf
SfSfSfSf
uuSf
uSfSfSfSSfSf
uuuS
kk
xi
kii
iiiii
iii
iuiuiui
kiu
iuiui
i
k
k
k
Monotone increasing
Why?Submodular!
Theorem
15
optimal. of)1(least at aluefunction v objective
ith solution wion approximatan produces which problem,on maximizati influencefor
algorithm time-polynomial a exists There
1 e
Exercise
16
ratio. eperformancion approximat theestimate Please
end );(\
;,..., subsets produce A to algorithm use begin do while
follows. asCover Set Min for ion approximatan design to
CoverageMax for ion approximat-)1( usemay We.1
1
1
1
k
k
SSXXSSk
X
e
Section 2.1-2.2
17
Independent System
• Consider a set E and a collection C of subsets of E. (E,C) is called an independent system if
CACBBA ,
Maximization
• c: E→R
max c(A) s.t. AεC • c(A) = ΣxεA c(x)
+
Greedy Approximation MAX
.output };{hen t
}{ if do to1for
;).()()(
ordering into in elements allSort 21
AxAA
CxAni
Axcxcxc
E
i
i
n
Theorem
.in
set t independen maximal of size minimum theis)( andin set t independen of size maximum theis )( where
)()(max
)(*)(
Then solution. optimalan be *Let Algorithm.Greedy by the obtained beLet
FFv
FFuFvFu
AcAc
AA
EFG
G
Proof
)(|*|))()((|*|
*)( Similarly,
)(||))()((||
|)||)(|(||)(
)( }. ..., ,{ Denote
1
1
1
1
1
1
2
111
1
nn
n
i
iii
nGn
n
i
iiGi
n
i
GiGiiG
G
ii
xcAExcxcAE
Ac
xcAExcxcAE
AEAExcAExc
AcxxE
. ofset t independen maximal a is is, that ),(||
show only to need We).(|*| that Note
.|||*|max
)(*)(
Therefore,
1
i
GiiGi
ii
Gi
i
niG
EAEEvAE
EuAEAEAE
AcAc
. choosingnot of rule theingcontradict t,independen is }{)( So, t.independen
is }{)(such that \ a is Then therein maximalnot is set t independen Suppose
1
.
j
jGj
jGiGij
iGi
xxAE
xAEAExEAE
THANK YOU!