Stochastic Finite Learning
Recently, we have developed a learning model, called stochastic finite learning, that makes a connection between concepts from PAC learning and inductive inference learning models. The motivation for this work is as follows. Within Gold’s (1967) model of learning in the limit many important learning problems can be formalized and it can be shown that they are algorithmically solvable in principle. However, since a limit learner is only supposed to converge, one never knows at any particular learning stage whether or not it has already been successful. Such an uncertainty may be not acceptable in many applications. The present paper surveys the new approach to overcome this uncertainty that potentially has a wide range of applicability.
KeywordsInductive inference average-case analysis stochastic finite learning conjunctive concepts pattern languages
Unable to display preview. Download preview PDF.
- 8.D. Haussler, Bias, version spaces and Valiant’s learning framework. “Proc. 8th National Conference on Artificial Intelligence” (pp. 564–569). Morgan Kaufmann, 1987.Google Scholar
- 10.M. Kearns L. Pitt, A polynomial-time algorithm for learning k-variable pattern languages from examples. “Proc. Second Annual ACM Workshop on Computational Learning Theory” (pp. 57–71). Morgan Kaufmann, 1989.Google Scholar
- 14.A. Mitchell, A. Sharma, T. Scheffer and F. Stephan, The VC-dimension of Subclasses of Pattern Languages, in “Proc. 10th International Conference on Algorithmic Learning Theory,” (O. Watanabe and T. Yokomori, Eds.), Lecture Notes in Artificial Intelligence, Vol. 1720, pp. 93–105, Springer-Verlag, Berlin, 1999.Google Scholar
- 15.L. Pitt, Inductive Inference, DFAs and Computational Complexity, in “Proc. 2nd Int. Workshop on Analogical and Inductive Inference” (K.P. Jantke, Ed.), Lecture Notes in Artificial Intelligence, Vol. 397, pp. 18–44, Springer-Verlag, Berlin, 1989.Google Scholar
- 16.R. Reischuk and T. Zeugmann, Learning One-Variable Pattern Languages in Linear Average Time, in “Proc. 11th Annual Conference on Computational Learning Theory-COLT’98,” July 24th–26th, Madison, pp. 198–208, ACM Press 1998.Google Scholar
- 17.R. Reischuk and T. Zeugmann, A Complete and Tight Average-Case Analysis of Learning Monomials, in “Proc. 16th International Symposium on Theoretical Aspects of Computer Science,” (C. Meinel and S. Tison, Eds.), Lecture Notes in Computer Science, Vol. 1563, pp. 414–423, Springer-Verlag, Berlin 1999.Google Scholar