i.i.d.

A fundamental assumption in statistical learning theory, stating that all samples are drawn independently from the same fixed but unknown distribution. The theoretical guarantees of many machine learning algorithms are built on the i.i.d. assumption.