Improvement on Boundary Searching of Accelerating Genetic Algorithm

Abstract

Accelerating Genetic Algorithm (AGA)'s disadvantages of unable to search the optimal solution when the solution is in the boundary of feasible region was proved through theoretical analysis and numerical experimentation. The solution of adding random individuals whose variable obeying to saddle distribution into initial population to increase the ability of searching the optimal solution in the boundary of AGA was proposed. The results of numerical tests show that the introduction of special individuals obeying to triangular distribution or pulse distribution can increase the accuracy by 3~10 times and convergence probability by 10%~20% when the global optimal solution is near the boundaries of feasible region and the accuracy is increased by 3~7 times and convergence probability by 25%~50% when global optimum is in the boundary of feasible region.

Predetermined time constant approximation method for optimising search space boundary by standard genetic algorithm

“…Search space boundary reduction for the candidate diameter for each link by pipe index vector and critical path method along with modified genetic operator's derivatives was proposed [8] [9]. Further, an improved AGM based on the saddle distribution by which adding random individuals into initial population to increase the searching ability of optimal solution was proposed [10]. Literature survey discloses that most techniques are considered based on limited or confined search space boundaries and involves complex mathematics.…”