Classification, Novelty Detection and Clustering for Point Pattern Data

Abstract

Point pattern data, also known as multiple instance data or bags, are abundant in nature and applications. However, machine learning problems for point patterns have not received much attention. In this work, we solve three fundamental machine learning problems, namely classification, novelty detection, and clustering, for point pattern data using two approaches: one with knowledge of the underlying data model (model-based approach), and one without (distance-based approach).