Which of the following statements is true about overfitting?

More training data may help prevent overfitting.

Generalization error is monotone with respect to the capacity$/$complexity of a model.

Overfitting may improve the generalization ability of a model.

If there is no noise, then there is no overfitting.