Computer networking representations for parallel distributed computing algorithms

Concerns the application of the mean field annealing (MFA) algorithm for combinatorial optimization problems. In particular, discrete optimization problems may be reduced to minimization of a 0-1 Hamiltonian. A significant algorithm for finding optimal solutions to such problems is the simulated annealing (SA) algorithm. It models the degrees of freedom in a problem as if it were a collection of atoms slowly being cooled into a ground state corresponding to the optimal solution to the problem. Although its principle may be based on complex and massive connections between network elements, there may be the bound of connective complexity in the realization. To overcome this problem we might apply the MFA, which is based on a stochastic optimization model. This paper proposes a network computing technique to perform the MFA, and shows how to apply it to the graph partitioning problem.