Georg Wölflein, Lucie Charlotte Magister, Pietro Liò, David J. Harrison, Ognjen Arandjelović
TL;DR提出了一种新的多实例学习（MIL）模型，使用距离感知自注意力（DAS-MIL）来考虑图像中不同补丁之间的相对空间关系，以提高医学图像分析的精度。通过在 MNIST-based MIL 数据集和公开的癌症转移检测数据集 CAMELYON16 上评估我们的模型，证明了该模型超越了现有的多数MIL方法。
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