Multi-Attention Multiple Instance Learning
This work addresses MIL challenges in domains like medical imaging by offering an incremental improvement in patch classification and interpretability.
The paper tackles the Multiple Instance Learning (MIL) problem by proposing MAMIL, a multi-attention method that considers neighboring patches to improve feature representation and classification accuracy for both patches and entire bags, with numerical experiments demonstrating its effectiveness.
A new multi-attention based method for solving the MIL problem (MAMIL), which takes into account the neighboring patches or instances of each analyzed patch in a bag, is proposed. In the method, one of the attention modules takes into account adjacent patches or instances, several attention modules are used to get a diverse feature representation of patches, and one attention module is used to unite different feature representations to provide an accurate classification of each patch (instance) and the whole bag. Due to MAMIL, a combined representation of patches and their neighbors in the form of embeddings of a small dimensionality for simple classification is realized. Moreover, different types of patches are efficiently processed, and a diverse feature representation of patches in a bag by using several attention modules is implemented. A simple approach for explaining the classification predictions of patches is proposed. Numerical experiments with various datasets illustrate the proposed method.