Search

Finding ways of recognizing global optimum is the very fundamental, unsolved problem in existing optimization theories. We can not establish a true theory of optimization without it. Also, it is very hard to construct effective algorithms for finding global optimum. This paper presented a new optimization principle, called cooperative optimization, for solving this extremely important problem in optimization theory. A number of global optimality conditions are provided in a general form. The application of cooperative optimization in coding yields near-perfect results in finding global optima, significantly better than the most powerful optimization algorithm ever found so far. [ABSTRACT FROM AUTHOR]

In this paper, we propose a general extension of constraint propagation for constraint optimization based on cooperative computation. It is similar both in principle and operations to constraint propagation. In principle, it eliminates variable values by checking the feasibility that they can be in any global optimum. In operations, it is based on parallel, local iterative computations. The proposed algorithm returns both a solution and a global lower bound at each iteration. As an approximation algorithm for optimization, it significantly outperform classical optimization methods, such as simulated annealing and local search with multi-restarts in practice. [ABSTRACT FROM AUTHOR]