how community structure aects epidemics - clara stegehuis
TRANSCRIPT
How community structure affects epidemicsClara Stegehuis
joint work with Remco van der Hofstad and Johan van Leeuwaarden
Eindhoven University of Technology
How does community structure affect epidemics?
Hierarchical configuration model
Epidemics
Configuration model [Bollobas 1980]
Configuration model [Bollobas 1980]
Configuration model [Bollobas 1980]
Configuration model [Bollobas 1980]
Community structure is lost!
Original network Configuration model
Hierarchical configuration model
Instead of pairing vertices, pair communities with half-edges
Start Possible result
Easy to analyze
Use that on community level still a configuration model to obtain:
I clustering
I component sizes
I degree correlations
.........
...
· · · · · · · · ·
......
Easy to analyze
Use that on community level still a configuration model to obtain:
I clustering
I component sizes
I degree correlations
.........
...
· · · · · · · · ·
......
HCM*
Randomize also edges within communities
Start Possible result
Different models
CM
Fixed degrees
HCM*
Fixed sets ofcommunities
HCM
Fixed communitystructure
Increasing randomness
Hierarchical configuration model
Epidemics
Bond percolation
Infect neighbor with probability p
S : fraction of vertices in largest component
Bond percolation
Infect neighbor with probability pS : fraction of vertices in largest component
Bond percolation results
Email network
0 0.2 0.4 0.6 0.8 10
0.2
0.4
0.6
0.8
1
edge occupation probability p
Siz
eof
larg
estc
ompo
nent
dataHCMHCM*CM
Internet router network
0 0.2 0.4 0.6 0.8 10
0.2
0.4
0.6
0.8
1
edge occupation probability pS
ize
ofla
rges
tcom
pone
nt
dataHCMHCM*CM
SIR epidemic
Email network
0 5 10 15 20 25 300.00
0.05
0.10
0.15
time
infe
cted
node
s
dataHCMHCM*CM
Internet router network
0 5 10 15 20 25 300.00
0.10
0.20
0.30
timein
fect
edno
des
dataHCMHCM*CM
Conclusion
I Communities are important for how epidemics spread acrossnetworks
I Exact community shapes are less important
I Communities may inhibit or enforce epidemics
Open problem
Overlapping communities