On the effective distribution and maintenance of knowledge represented by complementary graphs

Abstract

Graph transformations are a powerful tool enabling the formal description of the behavior of software systems. In most cases, however, this tool fails due to its low efficiency. This can be overcome by introducing parallel graph transformations. The concept of complementary graphs enables two things: the decomposition of a centralized graph into many cooperating subgraphs, and their parallel transformations. Such a model is very useful in an agent environment, where subgraphs represent an individual knowledge of particular agents; this knowledge may be partially replicated and exchanged between the agents. The rules of a cooperation and an implicit synchronization of a knowledge, represented in this way, have been already defined in [10]. The second very important issue is the way of an initial graph distribution assuming the size criterion: the heuristic method proposed previously succeeds in 60% (i.e. 60% of subgraphs is consistent with the criterion). The method presented in this paper gives over 90% fit. © 2012 Springer-Verlag Berlin Heidelberg.