Accuracy improvement of genetic algorithm for obtaining floating-point solution

The aim of our study is implementation of genetic algorithm (GA) in FPGA hardware. We use GA for obtaining floating-point solutions accurately. For this purpose, we propose applying a gray-coded floating-point format to GA to improve accuracy of the solutions. In this paper, we show the result of simulations using a gray-coded floating-point format. We evaluate performance of the proposed GA by obtaining solutions of five-dimensional Sphere function and two-dimensional Rosenbrock function. In these experimentations, we focused on mutation probability which is one of the parameters of GA for improving its accuracy. In the results, there was a trade-off between convergence speed and speed of finding the optimal solution depending on the mutation probability. However, we showed that our theory can obtain the optimal solutions effectively compared with the normal floating-point format.