We introduce, discuss, and study a model for inductive inference from samplings, formalizing an idea of learning different “projections” of languages. One set of our results addresses the problem of finding a uniform learner for all samplings of a language from a certain set when learners for particular samplings are available. Another set of results deals with extending learnability from a large natural set of samplings to larger sets. A number of open problems is formulated.
Jain, Sanjay and Efim Kinber. "Inductive Inference of Languages from Samplings." Proceedings. Algorithmic Learning Theory, 21st International Conference, ALT 2010, Canberra, Australia, October 6-8, 2010.