The "curse of dimensions" is a term that describes the many difficulties that arise in machine learning tasks as the number of features in the dataset increases. One way to solve this problem is to reduce the number of features to be provided to the model during the learning phase. This reduction in the number of dimensions can be done in two ways, either by merging dimensions together or by selecting a subset of dimensions. There are many methods to select the dimensions to be kept. One technique is to use a genetic algorithm to find a subset of dimensions that will maximize the accuracy of the classifier. A genetic algorithm specially created for this purpose is called genetic algorithm with aggressive mutation. This very efficient algorithm has several particularities compared to classical genetic algorithms. The main one is that its population is composed of a small number of individuals that are aggressively mutated. Our contribution consists in a modification of the algorithm. Indeed we propose a different version of the algorithm in which the number of mutated individuals is reduced in favor of a larger population. We have compared our method to the original one on 17 datasets, which allowed us to conclude that our method provides better results than the original algorithm while reducing the computation time. |

*** Title, author list and abstract as seen in the Camera-Ready version of the paper that was provided to Conference Committee. Small changes that may have occurred during processing by Springer may not appear in this window.