Parallel computing application to metaheuristic optimization algorithm

We need to take advantage of parallel computing's computational efficiency in improving the performance of metaheuristic optimization algorithm (MOA). Most previous studies have utilized parallel computing method only to evaluate a group of solutions fitness simultaneously. Although significantly reducing CPU computation time for the allowed number of function evaluation, this approach does not have any guarantees for improving MOA's robustness, i.e., ability to consistently find an global optimal solution. One way is to run multiple MOAs concurrently under different processors between which search information is broadcast and shared at a interval.

An interesting research topic is to compare the performance of two schemes: (1) conventional MOA running on regular computing scheme with N population and n function evaluations and (2) parallel MOA running on m processors with N/m population and n/m function evaluations.

<Word cloud from the link https://www.linkedin.com/pulse/parallel-computing-101-zile-rehman/>