Prudence and other conditions on formal language learning
Mark A. Fulk · Information and Computation · 1990
Inductive inference (IIMs) are used to model, among other things, human language learning. Various restrictions on the behavior of IIMs are investigated, the question of interest being whether restricted IIMs can be as powerful as unrestricted IIMs. It is shown that set-driven IIMs are limited in power, whereas order-independent, rearrangement-independent, and prudent IIMs are not. The motivation of formal language learning theory from human language learning is questioned.