Three-parent GA: A Global Optimization Algorithm
Amar Singh, Shakti Kumar, Ajay Singh and S.S. Walia
This paper proposes a new global optimization approach named three-parent genetic algorithm. The algorithm is inspired by the three-parent biological process that has been successfully employed in the field of genetics by the doctors to get rid of mitochondrial diseases which otherwise the offspring would have suffered. Thus, making three-parent children, healthier than the normal two-parent children. In the three-parent concept, the cell body with defective mitochondria of original mother is separated from nucleus and the cell body is destroyed. The separated nucleus is placed in the donors denucleared cell body containing healthy mitochondria. Thereafter the egg is fertilized. The process was observed to produce fitter offspring. We conceptualized this algorithm based upon the 3-parent genetic process. We implemented the algorithm in MATLAB, tested it on the CEC-2014 test bench and compared the performance with 16 other algorithms. The proposed algorithm outperformed all other algorithms on 7 test functions. None of the other algorithms could match this performance. In addition for functions f3 and f8 of CEC-2014 it gave the best performance but this performance was equaled by few other algorithms also. In total the proposed algorithm gave best performance on 9 functions. Further, the paper proposes a 3PGA based new dynamic optimal cost path evaluation approach in WMNs. The approach was implemented in MATLAB and the performance was compared with 8 other approaches. In this application also the proposed approach gave best performance of all the approaches for WMNs of size greater than 1000 nodes.
Keywords: Global optimization, GAs, 3PGA, WMNs, firefly, shortest path routing.