www-ai.cs.tu-dortmund.de/LEHRE/SEMINARE/SS09/AKTARBEITENDESDM/LITERATUR/CommunityDiscovery.pdf
LNAI 5212 - Hierarchical, Parameter-Free Community Discovery
intuition.
Let A1 := {0, 1, 0, 1, · · · , 0, 1}, with alternating values. We have C(A1|M (1) 1 ) =
16H(1/2) = 16 and C(M (1) 1 ) = log!16 + #log(16 + 1)$ = 10 + 5 = 15. However,
for M (2) 1 the best choice [...] Figures 2a and 3 follows such a structure, consisting of G1 = (I1, J1, A1,1), G2 = (I1, J2, A1,2), G3 = (I2, J1, A2,1), and G4 = (I2, J2, A2,2), where I1 is the set of “computer science researchers” and I2 the [...] choice is n1 = 15, with C(A1|M (2)
1 ) ' 15 and C(M (2) 1 ) ' 19.
The total code lengths are C(A1, M (1) 1 ) ' 16 + 15 = 31 and C(A1, M
(2) 1 ) '
15 + 19 = 34. Thus, based on total code length, the simpler …