Skip to Main Content
Many real-life problems can be formulated as numerical optimization of certain objective functions. However, often an objective function possesses numerous local optima, which could trap an algorithm from moving toward the desired global solution. Evolutionary algorithms (EAs) have emerged to enable global optimization; however, at the present stage, EAs are basically limited to solving small-scale problems due to the constraint of computational efficiency. To improve the search efficiency, this paper presents a stochastic genetic algorithm (StGA). A novel stochastic coding strategy is employed so that the search space is dynamically divided into regions using a stochastic method and explored region-by-region. In each region, a number of children are produced through random sampling, and the best child is chosen to represent the region. The variance values are decreased if at least one of five generated children results in improved fitness, otherwise, the variance values are increased. Experiments on 20 test functions of diverse complexities show that the StGA is able to find the near-optimal solution in all cases. Compared with several other algorithms, StGA achieves not only an improved accuracy, but also a considerable reduction of the computational effort. On average, the computational cost required by StGA is about one order less than the other algorithms. The StGA is also shown to be able to solve large-scale problems.