Comparative Study on Fitness Landscape Approximation with Fourier Transform

We propose to apply n dimensional discrete Fourier transform (DFT) to a fitness landscape, search an elite individual using obtained principal frequency component and accelerate evolutionary computation (EC) search. a comparative evaluation with our previous works is conducted using eight benchmark functions. the evaluation shows that our proposed approach can obtain the accurate fitness landscape than that with 1 dimensional DFT, and EC acceleration performance can be improved significantly. However, it needs more computational time in the process of conducting n dimensional DFT than that in 1 dimension. We also investigate the computational complexity of the two approaches and some related issues.