Dual types of hypotheses in inductive inference

From MaRDI portal





In inductive inference one natural way to detect that the actual hypothesis is incorrect consists in finding out some argument at which the hypothesis is defined and differs from the function to be identified. The present paper is interested in characterizing the powers of identification with convergently and divergently incorrect intermediate hypotheses. Section 3 defines the special version of temporarily conform identification, characterizing its power by comparing it with other types. Section 4 shows that for temporarily conform identification there are identifiable function classes which cannot be identified with convergently incorrect intermediate hypotheses. Section 5 considers the dual situation where none of the hypotheses differs convergently from the function to be identified. It is proved that the resulting identification type with divergently incorrect intermediate hypotheses is equivalent to monotonic identification, where the hypotheses have to be charged such that the corresponding subfunctions converge monotonically to the unknown function. Section 6 provides a hierarchy showing that for identification of all identifiable classes of recursive functions there are necessary a number of alternations of convergently and divergently incorrect intermediate hypotheses. For the entire collection see [Zbl 0819.68110].











This page was built for publication: Dual types of hypotheses in inductive inference

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4202920)