Input-Dependence in Function-Learning

In the standard model of inductive inference, a learner gets as input the graph of a function, and has to discover (in the limit) a program for the function. In this paper, we consider besides the graph also other modes of input such as the complement of the graph, the undergraph and the overgraph of the function. The relationships between these models are studied and a complete picture is obtained. Furthermore, these notions are also explored for learning with oracles, learning in teams and learning in the presence of additional information. Keywords. Inductive inference; recursion theory; various forms of input presentation; team learning; learning with additional information. Mathematical Subject Classification. 68Q32 (Computational Learning Theory); 03D20 (Recursive Functions).

• Type: Conference
• Year: 2007
• Where: CIE
• Authors: Sanjay Jain, Eric Martin, Frank Stephan
• Keywords: Additional Information, CIE 2007, Inductive Inference, Input