Coordination in Coarse-Grained Decomposition

The dominance of parallel computers in the supercomputer field provides great incentive for the development of parallel algorithms in mathematical programming. Fortunately, parallel computers provide a good medium for the implementation of many types of decomposition algorithms for large-scale optimization. In this paper, a variety of coordination methods are studied for coarse-grained decomposition techniques. Some are variants of existing approaches that are designed to enhance rates of convergence and/or parallel efficiency, while others are novel methods that try to take maximum advantage of the parallel environment.