An Effective Approach to Handling Noisy Domains

概要

論文の詳細を見る
In inductive logic programming, concept formation in noisy domains can be considered as learning from noisy data. This paper describes an approach to learning from noisy examples with an approximate theory. Although this kind of learning is realistic in practical domains, there are few systems which can deal with such problems. The proposed approach includes a theory preference criterion, which extends the minimum description length (MDL) principle by unifying model complexity and exception cost. Model complexity is the encoding cost for an algorithm to obtain a logic program; exception cost is the encoding length of the training examples misclassified by a theory. After revising all the clauses in the initial theory, the approach learns more accurate clauses by adopting heuristic constraints. Experiments show that our approach appears to be more accurate compared with existing approaches.
社団法人人工知能学会の論文
1998-03-01