this is my study note of some literature about LFR benchmark.
LFR benchmark is a artificial network in order to imitate real network,but LFR benchmark,actually in my opinion,is not only a network,also a measure(procedure) creating one.
so,in this note.I mainly introduce the procedure of construction of LFR benchmark graphs.
Ⅰ.THE FEATURE OF LFR BENCHMARK
what is a real network?there are someindividual and relationship.the node of graph is individual,and the link of graph is relationship,respectively.
node: the one of foundation of graph.
link: the feature of graph.the node is isolated if it has no link to others.
In order to make the description of real network more specific,one introduce two extra parameters.
degree: the sum of links of a node.
internal degree: the sum of links of a node connecting the node of lts community.
external degree: the sum of links of a node connecting the node of other community.
ps.In the benchmark,we need to set every degrees obey a power law distribution.
community size: the sum of nodes of a group.
ps1.we also need to set every community size obey a power law distribution.
ps2.the community is the set of expensive nodes and links.
Ⅱ.THE PROCEDURE OF CONSTRUCTION OF LFR BENCHMARK
we assume that both the degree and the community size distribution are power laws,with exponents γ and β,respectively,The number of nodes is N,the average degree is <k>[1].there is a flow char of the procedure of construction of LFR benchmark graphs.
ps.for the exponents we have taken typical values of real network: 2≤γ≤3,1≤β≤2.[1]
[1]Andrea L, Santo F, Filippo R. Benchmark graphs for testing community detection algorithms.[J]. Physical Review E, 2008, 78(4):561-570.
[2]Andrea L, Santo F. Community detection algorithms: a comparative analysis.[J]. Physical Review E Statistical Nonlinear & Soft Matter Physics, 2009, 80(5 Pt 2):2142-2152.