The NeuralNet function uses backpropagation to train neural networks. You must provide input data and argument settings for training the networks; the function creates the fitted weights of the neural network. The NeuralNet function is optimized for performance on very large datasets (millions of rows).