Abstract
For 50 years, research in the area of inductive inference aims atinvestigating the learning of formal languages and is influenced bycomputability theory, complexity theory, cognitive science, machine learning,and more generally artificial intelligence. Being one of the pioneers, Goldinvestigated the most common formalization, learning in the limit both fromsolely positive examples as well as from positive and negative information. Thefirst mode of presentation has been studied extensively, including insights inhow different additional requirements on the hypothesis sequence of the learneror requested properties of the latter itself, restrict what collections oflanguages are learnable. We focus on the second paradigm, learning from informants, and study howimposing different restrictions on the learning process effects learnability.For example, we show that learners can be assumed to only change theirhypothesis in case it is inconsistent with the data (such learners are calledconservative). Further, we give a picture of how the most important learningrestrictions relate. Our investigations underpin the claim for delayabilitybeing the right structural property to gain a deeper understanding concerningthe nature of learning restrictions.