[dekhn]

Learning is using observations to create/update a model that makes predictions which are more accurate than chance. At some point the model ends up having generalizability beyond the domain.