Quantifying inductive bias: AI learning algorithms and Valiant's learning framework
Artificial Intelligence
Machine Learning
Machine Learning
A Lattice Machine Approach to Automated Casebase Design: Marrying Lazy and Eager Learning
IJCAI '99 Proceedings of the Sixteenth International Joint Conference on Artificial Intelligence
Hi-index | 0.00 |
A version space is a set of all hypotheses consistent with a given set of training examples, delimited by the specific boundary and the general boundary. In existing studies [4, 5, 3] a hypothesis is a conjunction of attribute-value pairs, which is shown to have limited expressive power [6].In this paper we investigate version space in a more expressive hypothesis space, where a hypothesis is a hyperrelation, which is in effect a disjunction of conjunctions of disjunctions of attribute-value pairs. We propose to use an inductive bias, E-set, which turns our attention to equilabelled, supported, and maximal hypertuples. We characterise version space in such a hypothesis space under this bias and show the relationship between the specific boundary and general boundary with respect to unequivocal data, a special subset of the data space. We present experimental results on some public datasets.