Learning to Classify Incomplete Examples

Most research on supervised learning assumes the attributes of training and test examples are completely speciied. Real-world data, however, is often incomplete. This paper studies the task of learning to classify incomplete test examples, given incomplete (resp., complete) training data. We rst show that the performance task of classifying incomplete examples requires the use of default classiication functions which demonstrate nonmonotonic classiica-tion behavior. We then extend the standard pac-learning model to allow attribute values to be hidden from the classiier, investigate the robustness of various learning strategies, and study the sample complexity of learning classes of default classiication functions from examples.