The practice of trying to use the least data possible for training a Model without compromising performance.

Methods