In general, parameters Ɵ reflecting relative importance of the configurations, are unknown, but may be estimated from data. In principle, every clique may contribute in its own magnitude, which would have result, however, in too many parameters to estimate. Parsimony concerns, thus, encourage researchers to apply homogeneity assumption. This means that configurations of a specific type are assumed to contribute equivalently into the network’s probability. This also can be interpreted as that structurally equivalent networks, that is networks that have equal counts of relevant configurations, are equal in probability. Conditions under which networks are classified as isomorphic stem from dependence assumptions. In a sense, homogeneity assumption divides configurations into classes of equivalence (Robins & Wasserman, 2005) according with effect they create on the network’s probability. This allows not to think of each network feature, but switch to counts of them as they become sufficient statistics, with which network’s probability can be fully determined (Koskinen & Snijders, 2012).