On the error correcting power of pluralism in BC-type inductive inference (Q1068540)

The author considers \(BC^ m\) inference (introduced by J. Bārzdiņš and J. Case, independently), showing that it is always possible to reduce errors for some forms of inductive inference by increasing the number of machines involved in the inference process. Let M be an inductive inference machine and f a total function; the machine \(MBC^ m\) identifies the function f (one writes \(f\in BC^ m(M))\) if and only if, when M is successively fed the graph of f as input, it outputs over time a sequence of programs \(P_ 0,P_ 1,..\). such that \((\forall^{\infty}i)\phi P_ i=^ mf\), where \(g=^{\infty}f\) means that g and f disagree at at most m places. If, for some integer \(k\leq m\), only finitely many of the programs \(P_ i\) in the sequence above have more than k ''errors of commission'' then one writes \(f\in BC^{m:k}(M)\). Some classes of functions are defined and, by using relations between these classes, the results announced are proved.