This paper considers the notion of identifying subsets of critical data instances in data sets. Critical nuggets of information can take the following form during classification tasks: small subsets of data instances that lie very close to the class boundary and are sensitive to small changes in attribute values, such that these small changes result […]