A machine learning algorithm is a computational procedure that determines or updates the parameters of a machine learning model from data according to defined criteria. It specifies how the model learns from examples, errors, rewards, or other signals.