1 maximal independent set. 2 independent set (is): in a graph, any set of nodes that are not...
Post on 18-Dec-2015
217 Views
Preview:
TRANSCRIPT
1
Maximal Independent Set
2
Independent Set (IS):
In a graph, any set of nodes that are not adjacent
3
Maximal Independent Set (MIS):
An independent set that is nosubset of any other independent set
4
Applications in Distributed Systems
•In a network graph consisting of nodes representing processors, a MIS defines a set of processors which can operate in parallel without interference
•For instance, in wireless ad hoc networks, to avoid interference, a conflict graph is built, and a MIS on that defines a clustering of the nodes enabling efficient routing
5
A Sequential Greedy algorithm
Suppose that will hold the final MISI
Initially I
G
6
Pick a node and add it to I1v
1v
Phase 1:
1GG
7
Remove and neighbors )( 1vN1v
1G
8
Remove and neighbors )( 1vN1v
2G
9
Pick a node and add it to I2v
2v
Phase 2:
2G
10
2v
Remove and neighbors )( 2vN2v
2G
11
Remove and neighbors )( 2vN2v
3G
12
Repeat until all nodes are removed
Phases 3,4,5,…:
3G
13
Repeat until all nodes are removed
No remaining nodes
Phases 3,4,5,…,x:
1xG
14
At the end, set will be an MIS of I G
G
15
Worst case graph (for number of phases):
n nodes
Running time of algorithm: m)O(n
Number of phases of the algorithm: )(nO
16
A General Algorithm For Computing MIS
Same as the sequential greedy algorithm,but at each phase we may select any independent set (instead of a single node)
17
Suppose that will hold the final MISI
Initially I
Example:
G
18
Find any independent set 1I
Phase 1:
And insert to :1I I 1III
1GG
19
1I )( 1INremove and neighbors
1G
20
remove and neighbors 1I )( 1IN
1G
21
remove and neighbors 1I )( 1IN
2G
22
Phase 2:
Find any independent set 2I
And insert to :2I I 2III
On new graph
2G
23
remove and neighbors 2I )( 2IN
2G
24
remove and neighbors 2I )( 2IN
3G
25
Phase 3:
Find any independent set 3I
And insert to :3I I 3III
On new graph
3G
26
remove and neighbors 3I )( 3IN
3G
27
remove and neighbors 3I )( 3IN
No nodes are left
4G
28
Final MIS I
G
29
The number of phases depends on the choice of independent set in each phase:
The larger the independent set at eachphase the faster the algorithm
Observation:
30
Example:If is MIS, 1 phase is needed
1I
Example:If each contains one node, phases are needed
kI)(nO
(sequential greedy algorithm)
31
A Randomized Sync. Distributed Algorithm
Same with the general MIS algorithm
At each phase the independent setis chosen randomly so that it includesmany nodes of the remaining graph
32
Let be the maximum node degree in the whole graph
d
1 2 d
Suppose that is known to all the nodesd
33
Elected nodes are candidates forindependent set
Each node elects itself with probability
At each phase :k
kI
dp
1
1 2 d
kGz
z
34
However, it is possible that neighbor nodes may be elected simultaneously
Problematic nodeskG
35
All the problematic nodes must be un-elected. The remaining elected nodes formindependent set kI
kGkI
kIkI
kI
36
Success for a node in phase : disappears at end of phase(enters or )
Analysis:
kGz
kI
1 2 y
No neighbor elects itself
z
z
k)( kIN
k
A good scenariothat guaranteessuccess
elects itself
37
Basics of Probability
E: finite universe of events; let A and B denote two events in E; then:
1. A B is the event that either A or B occurs;
2. A B is the event that both A and B occur.
38
Probability of success in phase:
1 2 y
p
p1
p1p1
dy pppp )1(1
z
No neighbor should elect itself
At least
elects itself
39
Fundamental inequalities
tn
t ent
nt
e
11
21n
nt ||
10 p
1k
k
kp
p
11
40
Probability of success in phase:
ed
ded
dd
ppppd
dy
2
1
11
1
11
1
)1(1
At least
For 2d
First ineq. with t =-1
41
Therefore, node will enterand disappear at the end of phasewith probability at least
1 2 y
z
z kI
ed21
k
42
Expected number of phases until nodedisappears:
at most ed2phase in success of yprobabilit
1 phases
z
43
after phases
Bad event for node :
ned ln4
node did not disappear
Probability (First ineq. with t =-1 and n=2ed):
2ln2
ln411
21
1need n
ned
z
z
44
after phases
Bad event for any node in :
ned ln4
at least one node did not disappear
Probability:
nnnx
Gx
11) for event bad of ty(probabili 2
G
45
within phases
Good event for all nodes in :
ned ln4
all nodes disappear
Probability:
n1
-1event] bad of yprobabilit[1
(high probability)
G
46
Total number of phases:
)log(ln4 ndOned
Time duration of each phase: )1(O
Total time: )log( ndO
with high probability
47
Luby’s MIS Distributed Algorithm
Runs in time in expected case)(lognO
)log(log ndO with high probability
this algorithm is asymptoticallybetter than the previous
48
Let be the degree of node)(vd
1 2 )(vd
v
v
49
Each node elects itselfwith probability
At each phase :k
kI
)(21
)(vd
vp
kGv
1 2 )(vd
v
degree ofin
Elected nodes are candidates for theindependent set
v
kG
50
)(vd
v
z
)(zd
If two neighbors are elected simultaneously,then the higher degree node wins
Example:
)(vd
v
z
)(zd
12
1
2
12
1
2)()( vdzd
if
51
)(vd
v
z
)(zd
If both have the same degree, ties are broken arbitrarily
Example:
)(vd
v
z
)(zd
12
1
2
12
1
2)()( vdzd
if
52
Problematic nodeskG
Using previous rules, problematic nodesare removed
53
kG
The remaining elected nodes formindependent set kI
kI
kIkI
kI kI
kI
54
at least one neighbor enters
Analysis
2z
A good event for node
1 2)(vd
v
1z)(vdz
kI:vH
Consider phasek
v
55
1z)(vdz
At end of phasek
If is true, then andwill disappear at end of current phase
v)( kINv vH
56
At least one neighbor ofelects itself with probability at least
2z
1 2)(vd
v
1z)(vdz
LEMMA 1: v
)(~
2
)(
1 vd
vd
e
)(max)(~
)(zdvd
vNz maximum neighbor degree
57
2z
1 2)(vd
v
1z)(vdz
No neighbor of elects itself with probability
)(
))(1(vNz
zp
vPROOF:
(the elections are independent)
58
)(~
2
)()(
~2
)()(~
2)(
)(~
2
11
)(~
2
11 vd
vdvd
vdvdvd
evdvd
)()( )(21
1))(1(vNzvNz zd
zp
maximum neighbor degree)(max)(~
)(zdvd
vNz
59
2z
1 2)(vd
v
1z)(vdz
Therefore, at least one neighbor of Elects itself with probability at least
v
)(~
2
)(
1 vd
vd
e
END OF PROOF
60
If a node elects itself,then it enters with probabilityat least
z
1 2
)(zdz
1 2
)(zdzv v
LEMMA 2:
kI
21
61
1 2
)(zdz
)()( 1 zdud
1u
u)()( zdud
Node enters if no neighbor of same or higher degree elect itself
zkI
PROOF:
62
1 2
)(zdz
Probability that some neighbor of with same or higher degree elects itself
)()( 1 zdud
1u
u
neighbors of same or higher degree
2
1
)(2
)(
)(2)(
itself)] elects (node[
1i
k
zd
zd
zdup
uP
i
k
)()( zdud
z
63
Probability that that no neighbor ofwith same or higher degree elects itself
21
21
-1itself)] elects (nodeP[-1
itself)] elects node (no[
k
k
kk
u
uP
neighbors of same or higher degree
z
1 2
)(zdz
)()( 1 zdud
1u
u
)()( zdud
64
1 2
)(zdz
Thus, if elects itself, it enterswith probability at least
z kI
21
1 2
)(zdzv
v
END OF PROOF
65
2z
1 2)(vd
v
1z)(vdz
at least one neighbor of enters kI:vH v
)(
~2
)(
121
][ vd
vd
v eHPLEMMA 3:
66
and no node is elected
neighbor is iniz kI
121 ,,, izzz
2z
12
)(vdv
1z)(vdz
New event
iz1iz kI
1ii
:iY
PROOF:
67
)(21 ,,, vdYYY The events
are mutually exclusive
)(
1)(1][
vd
iii
vdiYPYP
68
i
vdiv YPHP
)(1][
It holds:
)(
1)(1][][
vd
iii
vdiv YPYPHP
Therefore:
69
and no node elects itself
elects itselfiz
kI
121 ,,, izzz
2z
12
)(vdv
1z)(vdz
iz1iz
1ii
:iA
][][][ iii BPAPYP
:iB izafter elects itself, it enters
1u 2u ku
70
21
][ iBP (from Lemma 2)
2z
12
)(vdv
1z)(vdz
iz1iz
1ii
2z
12
)(vdv
1z)(vdz
iz1iz kI
1ii
1u 2u ku1u 2u ku
kI:iB izafter elects itself, it enters
71
)(
1
)(
1
)(
1
][21
][][][][vd
ii
vd
iii
vd
iiv APBPAPYPHP
21
][ iBP
72
and no node elects itself
elects itselfiz
121 ,,, izzz :iA
)(
1d(v)i1][]P[
elected] is of neighbor one least at[vd
iii APA
vP
The events are mutually exclusive)(21 ,,, vdAAA
73
)(~
2
)(
1elected] is of neighbor one least at[ vd
vd
evP
We showed earlier (Lemma 1) that:
Therefore:
)(~
2
)()(
1
1][ vd
vdvd
ii eAP
74
)(
~2
)()(
1
121
][21
][ vd
vdvd
iiv eAPHP
Therefore node disappears in phasewith probability at least
v k
END OF PROOF
75
Let be the maximum node degree in the graph
kd
kG
Suppose that in : 2
)( kdvd
Then, )(2)(~
vdvd
ceeHP vd
vd
v
41
)(~
2
)(
121
121
][constant
kG
76
(thus, nodes with high degreewill disappear fast)
2)( kd
vd a node with degree
with probability at least c
Thus, in phasek
disappears
77
Suppose that the degree of remains at least for the next phases
Consider a node which in initial graphhas degree
2)(
dvd
Gv
v
2d
Node does not disappear within phases with probability at most
v
)1( c
78
nc
1log3 1Take
Node does not disappear within phases with probability at most
v
3
1log3 1
)1()1( 1
ncc nc
79
Thus, within phases nc
1log3 1
v either disappears or its degree gets less than
with probability at least
2d
3
11
n
80
by the end of phases nc
1log3 1
there is no node of degree higher than2d
with probability at least (ineq. 2)
Therefore,
23
11
11
nn
n
81
In every phases,nc
1log3 1
the maximum degree of the graphreduces by at least half, with probability at least
2
11
n
82
)log(log1
log3log 1 ndOn
d c
Maximum number of phases until degreedrops to 0 (MIS has formed)
with probability at least (ineq. 2)
nnn
nd 11
11
11
22
log
top related