Solutions to Parallel and Distributed Computing Problems: Lessons from Biological Sciences

3.5 SUMMARY

A taxonomy for hybrid biologically inspired metaheuristics has been presented. It considers solutions to design and implementation issues. Treating the two problems orthogonally is beneficial because it allows one to study, understand, classify, and evaluate the algorithms using a well-defined set of criteria.

A high percentage of metaheuristics hybridizing population-based metaheuristics with local search heuristics has been proposed for various optimization problems. Pure population-based heuristics, such as genetic algorithms, genetic programming, evolution strategies, and ant colonies, are not well suited to a fine-tuned search in highly combinatorial spaces.

Most of the hybrid algorithms are sequential. The authors of those sequential approaches always indicate in their future work the parallelization of the algorithms. This is an indication of the growing interest in developing parallel hybrid algorithms. Parallel schemes ideally provides novel ways to parallelize hybrid algorithms by providing parallel models of the algorithms. Hence, instead of merely parallelizing and finely tuning a sequential algorithm that has important, but limited capabilities to be parallelized, parallel hybrids are inherently suited to parallel computer environments. Furthermore, it should be pointed out and emphasized that the HCH proposes natural ways to efficiently implement algorithms on heterogeneous computer environments, which is recognized as a very tough problem today.