# Evaluation of Cuckoo Search, Genetic Algorithm and Particle Swarm Optimization Algorithms on Benchmark Functions

Four classical benchmark functions namely Ackley (functionf_{1}), Rosenbrock (functionf_{2}), Rastrigin (function f_{3}) and Griewank (functionf_{4}; Table 7.2) were selected to compare the performance of the CS algorithm with the GA and the PSO algorithm. The description of the benchmark functions used in this work is listed in Table 7.2. The Rosenbrock function (functionf_{2}) is uni-modal, having only one minimum. The others are multi-modal, with a considerable number of local minima in the search space. All functions except *f*_{2} have their global minimum at the origin. For Rosenbrock function (f_{2}), the global minimum is at (1,1,___,1). Maximum number of iterations for the dimensions of

10, 20 and 30, were 500, 750 and 1000, respectively (Table 7.3). Other algorithm parameters are listed in Table 7.3. For each function, 20 trials were performed. The optimization algorithm has been coded in MATLAB® 2014b. The mean and the standard deviation obtained with the CS algorithm for each function are reported in Table 7.4 along with GA and PSO results which were taken from the literature (Karaboga & Basturk, 2007). CS outperformed GA and PSO algorithms while evaluating the function *f*_{2} whereas underperformed for function f_{3}. In case of function f_{1} and f_{4}, the performance of CS is equivalent to GA and PSO. Therefore, the CS algorithm has the ability to get out of the local minimum and can be efficiently used for the optimization of multi-variable and multi-modal functions.

FIGURE 7.1

Pseudo-code of CS algorithm.