With supervised learning, an artificial neural network learns to approximately represent the desired outputs (or targets) corresponding to particular specified input patterns.