Certainty Pooling for Multiple Instance Learning

TL;DR

This paper introduces Certainty Pooling, a new pooling operator for Multiple Instance Learning that improves robustness and explainability by incorporating model certainty, especially effective with small training datasets.

Contribution

The paper proposes Certainty Pooling, a novel permutation invariant pooling operator that enhances bag and instance predictions in Multiple Instance Learning.

Findings

Outperforms existing pooling methods on MNIST with low evidence ratio bags.

Achieves superior results on the Camelyon16 histopathology dataset.

More effective with small training sets.

Abstract

Multiple Instance Learning is a form of weakly supervised learning in which the data is arranged in sets of instances called bags with one label assigned per bag. The bag level class prediction is derived from the multiple instances through application of a permutation invariant pooling operator on instance predictions or embeddings. We present a novel pooling operator called \textbf{Certainty Pooling} which incorporates the model certainty into bag predictions resulting in a more robust and explainable model. We compare our proposed method with other pooling operators in controlled experiments with low evidence ratio bags based on MNIST, as well as on a real life histopathology dataset - Camelyon16. Our method outperforms other methods in both bag level and instance level prediction, especially when only small training sets are available. We discuss the rationale behind our approach and the reasons for its superiority for these types of datasets.