OPTIMAL MAPPING OF GRAPH HOMOMORPHISM ONTO SELF-ORGANIZING HOPFIELD NETWORK

In a recent paper by us, a novel programming strategy was proposed to obtain homomorphic graph matching using the Hopfield network. Subseque ntly a self-organisation scheme was also proposed to adaptively learn the constraint parameter which is required to generate the desired hom omorphic mapping for every pair of model and scene data. In this paper , an augmented weighted model attributed relational graph (WARG) repre sentation scheme is proposed. The representation scheme incorporates a distinct weighting factor and tolerance parameter for every model att ribute. To estimate the parameters in a simplified form of the model W ARG representation, learning schemes are presented. A heuristic learni ng scheme is employed to estimate suitable values for threshold parame ters. The computation of weighting factors is formulated as an optimis ation problem and solved using the quadratic programming algorithm. Th e formulation implicitly evaluates ambiguity, robustness and discrimin atory power of the relational attributes chosen for graph matching and assigns weighting factors appropriately to the chosen attributes. Exp erimental results are presented to demonstrate that the parameter lear ning schemes are essential when the models have intra-model ambiguity and the optimal set of parameters always generates a better mapping. ( C) 1997 Elsevier Science B.V.