Invariant classification algorithms satisfying non-redundancy and monotonicity conditions

Abstract Invariant classification algorithms that satisfy additional conditions of non-redundancy and monotonicity are considered. The equivalence of both types of conditions is proved for invariant algorithms. It is shown that correct invariant classification algorithms satisfying non-redundancy (monotonicity) conditions are generalizations of the KORA algorithm for general (not necessarily binary) features. An implementation of the algorithms is briefly described.