{Obtaining sets of training patterns; Bootstrap methods, e.g. bagging, boosting characterised by the incorporation of unlabelled data, e.g. multiple instance learning [MIL], semi-supervised techniques using expectation-maximisation [EM] or naïve labelling }