Multilayer perceptrons (MLP) models are suitable for both regression and classification tasks. In order to go from regression to classification, the only change that needs to be applied to the architecture is to adjust the activation used in the output layer so that the network predicts something within the appropriate range of the target labels. It would also be necessary to change the cost function to something appropriate to the context. In both settings, the process consists of the alternating steps of forward and backward propagation until the parameters of the network produce the most accurate output.