Learning Maximally Predictive Prototypes in Multiple Instance Learning

TL;DR

This paper introduces a simple, interpretable model that generates maximally predictive prototypes for multiple instance learning, improving classification accuracy and providing insights into the data.

Contribution

The paper presents a novel prototype-based framework for multiple instance learning that enhances interpretability and classification performance.

Findings

Achieves high accuracy on classical MIL benchmarks.

Provides interpretable prototypes that reveal data structure.

Efficiently maps bags to a distance feature space.

Abstract

In this work, we propose a simple model that provides permutation invariant maximally predictive prototype generator from a given dataset, which leads to interpretability of the solution and concrete insights to the nature and the solution of a problem. Our aim is to find out prototypes in the feature space to map the collection of instances (i.e. bags) to a distance feature space and simultaneously learn a linear classifier for multiple instance learning (MIL). Our experiments on classical MIL benchmark datasets demonstrate that proposed framework is an accurate and efficient classifier compared to the existing approaches.