LEARNING APPROACH FOR THE COMPUTATION OF GENERALIZED WHELAND POLYNOMIALS OF CHEMICAL GRAPHS

The generalized Wheland polynomial of a graph is computed as perfect m atchings of the associated edge-weighted complete graph. A depth-first -search algorithm with unit penetrance is developed for an effective c ombinatorial exploration of all perfect matchings of the edge-weighted complete graph, using a priori control knowledge. Since this is a typ ical NP-class problem, dynamic learning techniques are designed to pus h the tractable limit of the problem further. The choice of the learni ng technique depends solely on the amount of memory available for use since demand for memory increases concomitantly with knowledge acquisi tion. The skeleton of the algorithm is presented in a Pascal-like pseu do language.