Abstract:Traditional supervised learning tasks require a label for every instance in the training set, but in many real-world applications, labels are only available for collections (bags) of instances. This problem setting, known as multiple instance learning (MIL), is particularly relevant in the medical domain, where high-resolution images are split into smaller patches, but labels apply to the image as a whole. Recent MIL models are able to capture correspondences between patches by employing self-attention, allowing them to weigh each patch differently based on all other patches in the bag. However, these approaches still do not consider the relative spatial relationships between patches within the larger image, which is especially important in computational pathology. To this end, we introduce a novel MIL model with distance-aware self-attention (DAS-MIL), which explicitly takes into account relative spatial information when modelling the interactions between patches. Unlike existing relative position representations for self-attention which are discrete, our approach introduces continuous distance-dependent terms into the computation of the attention weights, and is the first to apply relative position representations in the context of MIL. We evaluate our model on a custom MNIST-based MIL dataset that requires the consideration of relative spatial information, as well as on CAMELYON16, a publicly available cancer metastasis detection dataset, where we achieve a test AUROC score of 0.91. On both datasets, our model outperforms existing MIL approaches that employ absolute positional encodings, as well as existing relative position representation schemes applied to MIL. Our code is available at https://anonymous.4open.science/r/das-mil.

Nested Multiple Instance Learning with Attention Mechanisms

Multiple-Instance Learning from Pairwise Comparison Bags

Address Instance-level Label Prediction in Multiple Instance Learning

Dual-stream Maximum Self-attention Multi-instance Learning

Deep Multiple Instance Learning with Distance-Aware Self-Attention

A New Multiple Instance Algorithm Using Structural Information.

Rethinking Multiple Instance Learning: Developing an Instance-Level Classifier via Weakly-Supervised Self-Training

A Multiclass Multiple Instance Learning Method with Exact Likelihood

Multi-Instance Learning with One Side Label Noise

Multi-head Attention-based Deep Multiple Instance Learning

Deep Multiple Instance Learning for Zero-Shot Image Tagging

Multi-Instance Learning with Any Hypothesis Class

Attention-Challenging Multiple Instance Learning for Whole Slide Image Classification

Multi-instance Positive and Unlabeled Learning with Bi-Level Embedding

Multiple instance learning from similarity-confidence bags

Smooth Attention for Deep Multiple Instance Learning: Application to CT Intracranial Hemorrhage Detection

Multiple instance learning: A survey of problem characteristics and applications

Multi-Instance Learning with Key Instance Shift

Instance-level Semisupervised Multiple Instance Learning.

Multi-Instance Learning with Emerging Novel Class

Multiple Instance Learning Framework with Masked Hard Instance Mining for Whole Slide Image Classification