Why might cross$-$validation be used instead of a single train$/$test split? To simplify the model training process To reduce the variability of the model evaluation To increase the training speed To ensure the model overfits the training data