Nested Multiple Instance Learning with Attention Mechanisms
This addresses a limitation in multiple instance learning for applications requiring nested structures, offering a method to predict latent labels, though it appears incremental as it builds on existing MIL with attention mechanisms.
The paper tackles the problem of weakly supervised learning in complex scenarios where labels are only available at a bag-of-bags level, proposing a Nested Multiple Instance with Attention (NMIA) model that performs comparably to conventional MIL in simple cases and provides insights into latent labels at different levels in complex settings.
Strongly supervised learning requires detailed knowledge of truth labels at instance levels, and in many machine learning applications this is a major drawback. Multiple instance learning (MIL) is a popular weakly supervised learning method where truth labels are not available at instance level, but only at bag-of-instances level. However, sometimes the nature of the problem requires a more complex description, where a nested architecture of bag-of-bags at different levels can capture underlying relationships, like similar instances grouped together. Predicting the latent labels of instances or inner-bags might be as important as predicting the final bag-of-bags label but is lost in a straightforward nested setting. We propose a Nested Multiple Instance with Attention (NMIA) model architecture combining the concept of nesting with attention mechanisms. We show that NMIA performs as conventional MIL in simple scenarios and can grasp a complex scenario providing insights to the latent labels at different levels.