Multiple Instance Learning with the Optimal Sub-Pattern Assignment Metric

TL;DR

This paper introduces the Optimal Sub-Pattern Assignment metric for multiple instance learning, enabling flexible solutions across clustering, classification, and novelty detection tasks with demonstrated effectiveness on various datasets.

Contribution

The paper presents a novel metric tailored for multiple instance learning, enhancing versatility in handling different learning tasks.

Findings

Effective in clustering, classification, and novelty detection

Demonstrated versatility on simulated and real datasets

Improves upon existing set distance measures

Abstract

Multiple instance data are sets or multi-sets of unordered elements. Using metrics or distances for sets, we propose an approach to several multiple instance learning tasks, such as clustering (unsupervised learning), classification (supervised learning), and novelty detection (semi-supervised learning). In particular, we introduce the Optimal Sub-Pattern Assignment metric to multiple instance learning so as to provide versatile design choices. Numerical experiments on both simulated and real data are presented to illustrate the versatility of the proposed solution.