Specifically, in the case of Neural Networks Algorithms, feature scaling benefits optimization by: It makes the training faster. It prevents the optimization from getting stuck in local optima. It gives a better error surface shape.