Random networks in ERGO format
These are the 4 randomly generated with Jaap Suermondt's code networks I used in
the paper on parallelization. The
code first builds a completely connected graph and then removes edges at random
to satisfy specified number of edges per node. These networks are quite
different from the used in practice and are very hard on the standard
Lauritzen-Spiegelhalter algorithm:
And these are the TL8 and MC8 networks I describe in the paper:
And these are BN2O networks I used in the paper on the similarity of
states. In the first, coefficients are generated with beta
distribution, and in the second the coefficients are taken
from the full CPCS network at random:
The CPCS (sub)networks are not given due to copyright restrictions.