Boundary examples are taken from training data or generated and are at or near the boundary between classes. It can be useful to identify boundary examples during machine learning as they are places where mis-classifications are likely and hence may be more heavily weighted to ensure accurate boundary behaviour. Boundary examples can also be used as part of explainable AI in order to help a human understand classes created by an AI system.
Used in Chap. 21: page 359
Identifying central and boundary examples