Before applying kNN to a classification task, it is common practice to rescale the data using a technique like min-max normalization. What is the purpose of this step?