Upload
dane
View
29
Download
0
Embed Size (px)
DESCRIPTION
Optimal Determination of Source-destination Connectivity in Random Graphs. Luoyi Fu, Xinbing Wang, P. R. Kumar Dept. of Electronic Engineering Shanghai Jiao Tong University Dept. of Electrical & Computer Engineering Texas A&M University. N nodes Each edge exists with probability p - PowerPoint PPT Presentation
Citation preview
Optimal Determination of Source-Optimal Determination of Source-destinationdestination
Connectivity in Random Graphs Connectivity in Random Graphs
Luoyi Fu, Xinbing Wang, P. R. Kumar Dept. of Electronic EngineeringShanghai Jiao Tong University
Dept. of Electrical & Computer EngineeringTexas A&M University
Random Graph: G(n,p) ModelRandom Graph: G(n,p) Model N nodesN nodes Each edge exists with probability pEach edge exists with probability p Proposed by Gilbert in 1959Proposed by Gilbert in 1959 It can also be called ER graphIt can also be called ER graph
2/182/18
Are S and D Connected?Are S and D Connected? Goal: Determine whether S and D are connected or notGoal: Determine whether S and D are connected or not
As quickly as possibleAs quickly as possible
i.e., by testing the fewest expected number of edgesi.e., by testing the fewest expected number of edges
S
D
3/183/18
Determined S-D connectivity in 6 number of edgesDetermined S-D connectivity in 6 number of edges
By finding a pathBy finding a path
S
D
edges testededges tested
4/184/18
Determined S-D disconnectivity in 10 number of edgesDetermined S-D disconnectivity in 10 number of edges
By finding a cutBy finding a cut
S
D
edges testededges tested
5/185/18
SS DD SS DD
Sometimes, S and D may be connected.Sometimes, S and D may be connected. Sometimes, S and D may be disconnected.Sometimes, S and D may be disconnected.
Termination time may be random.Termination time may be random.
We want to determine whether S and D are connected or not We want to determine whether S and D are connected or not By either finding a Path or a CutBy either finding a Path or a Cut By testing the fewest number of edgesBy testing the fewest number of edges
Quickest discovery of an S-D route has not been studied before.Quickest discovery of an S-D route has not been studied before.
Finding a shortest path is not the goal here.Finding a shortest path is not the goal here. Finding the shortest path is a well studied problem.Finding the shortest path is a well studied problem.
6/186/18
Test the direct edge between S and DTest the direct edge between S and D
Test a potential edge between S and a randomly Test a potential edge between S and a randomly chosen nodechosen node Contract S and the node into a component if an Contract S and the node into a component if an
edge exists between themedge exists between them Test the direct edge between CTest the direct edge between CSS and D and D
2 potential edges between nodes and D2 potential edges between nodes and D 3 potential edges between nodes and C3 potential edges between nodes and CSS
Test an edge between D and a randomly chosen Test an edge between D and a randomly chosen nodenode 2 potential edges between node 2 and CS2 potential edges between node 2 and CS 1 potential edges between node 3 and CS1 potential edges between node 3 and CS Test the edge between node 2 and DTest the edge between node 2 and D
Similar rules in generalSimilar rules in general
The Optimal Policy: A Five-node ExampleThe Optimal Policy: A Five-node Example
11
33
22
SS DDCCSS
CCDD
7/187/18
The Optimal Policy: General CaseThe Optimal Policy: General Case
……
……
..
Rule 1Rule 1: : Test if edge exists between CTest if edge exists between CSS and C and CD.D.
Policy terminates if the edge exists.Policy terminates if the edge exists.
Rule 2: Rule 2: List all the paths connecting CList all the paths connecting CSS to C to CDD with the with the
minimum number of potential edges.minimum number of potential edges. Not CNot CSS-C-C11-C-C22-C-CDD
But CBut CSS-C-C11-C-CDD
Find Set M that contains the minimum potential Find Set M that contains the minimum potential Cut between CCut between CSS and C and CD.D.
Rule 3: Rule 3: Sharpen Rule 2 by specifying which particular Sharpen Rule 2 by specifying which particular
edge in M should be tested.edge in M should be tested. Test any edges in MTest any edges in M connecting Cconnecting CSS to C to C11..
CCSS CCDD
CC11
CC22
CCrrMM
1n
2n
rn
1 1 2max , , , rn n n n
8/188/18
Proof of Rule 1: Test If the Direct Edge ExistsProof of Rule 1: Test If the Direct Edge Exists
Testing the direct edge at the first step is better than testing Testing the direct edge at the first step is better than testing at the second step.at the second step.
badAgoodA
SS DD SS DD
SS DD SS DD
SS DD SS DD
SS DD SS DD
SS DD
SS DD
terminateterminate
Same Same probabilityprobability
9/189/18
Induction on the number of edges tested before the direct edge is Induction on the number of edges tested before the direct edge is testedtested
terminateterminate
Terminate Terminate one step one step earlier!earlier!
Proof of Rule 3Proof of Rule 3
Testing CTesting CSS-C-C11 edge is better than testing C edge is better than testing CSS-C-C22 edge. edge.
SS DD
CC11
CC22
1Dk
2Dk
1Sk
2Sk1 2D Dk k
10/1810/18
SS DD
11
rr
22
…
1n
2n
rn
…
1 1 2max , , , rn n n n
M
Proof of Rule 3Proof of Rule 3
Take the graph on the right as example.Take the graph on the right as example.
Two policies:Two policies:
badAgoodA
SS DD
CC11
SS DD
CC11
SS DD
CC11
SS DD
CC11
CC22 CC22
11
11 22
33
CC22 CC22
11/1811/18
Induction on the number of potential edges in Induction on the number of potential edges in the graph.the graph.
Proof of Rule 3Proof of Rule 3
Stochastically couple edges under AStochastically couple edges under Agoodgood and A and Abadbad..
goodA
badA
SS DD
11
22
SS DD
11
22
12/1812/18
goodA
badA
Terminates earlier!Terminates earlier!
Testing CS -C1 edge is better than testing C1-CD edge.
Stochastic coupling argument
Induction on the number of potential edges in the graph
Proof of Rule 2Proof of Rule 2
DD
CC11 12k11k
In the set MIn the set M
SS
13/1813/18
Proof of Rule 2Proof of Rule 2
goodA
badA
One step earlier!One step earlier!
14/1814/18
Phase TransitionPhase Transition
1000 nodes1000 nodes
P~0: 999 edges from SP~0: 999 edges from S
P~1: 1 edge to DP~1: 1 edge to D
Phase transition: take a long time (around 15000 steps) to testPhase transition: take a long time (around 15000 steps) to test
Our policy is optimal for all p!Our policy is optimal for all p!
15/1815/18
Extension to Slightly More General GraphsExtension to Slightly More General Graphs
Series graphsSeries graphs
Parallel graphsParallel graphs
SP graphsSP graphs
PS graphsPS graphs
Series of parallel of series (SPS) graphsSeries of parallel of series (SPS) graphs
Parallel of series of parallel (PSP) graphsParallel of series of parallel (PSP) graphs
16/1816/18
Concluding RemarksConcluding Remarks
Whether ER are connected graphs is very well studied topic.Whether ER are connected graphs is very well studied topic.
Quickly testing connectivity is not.Quickly testing connectivity is not. (Surprisingly)(Surprisingly)
We provide the optimal testing algorithm.We provide the optimal testing algorithm.
Optimal for all p.Optimal for all p.
17/1817/18
Thank you !Thank you !
18/1818/18