MixBag: Bag-Level Data Augmentation for Learning from Label Proportions

TL;DR

MixBag introduces a novel bag-level data augmentation technique for learning from label proportions, improving instance classification accuracy by leveraging augmented bags and confidence interval loss, with broad applicability and proven effectiveness.

Contribution

This paper presents the first bag-level data augmentation method for LLP, enhancing classifier performance and compatibility with existing LLP techniques.

Findings

Improved instance classification accuracy with MixBag.

Effective augmentation across different LLP methods.

Validated through extensive experiments.

Abstract

Learning from label proportions (LLP) is a promising weakly supervised learning problem. In LLP, a set of instances (bag) has label proportions, but no instance-level labels are given. LLP aims to train an instance-level classifier by using the label proportions of the bag. In this paper, we propose a bag-level data augmentation method for LLP called MixBag, based on the key observation from our preliminary experiments; that the instance-level classification accuracy improves as the number of labeled bags increases even though the total number of instances is fixed. We also propose a confidence interval loss designed based on statistical theory to use the augmented bags effectively. To the best of our knowledge, this is the first attempt to propose bag-level data augmentation for LLP. The advantage of MixBag is that it can be applied to instance-level data augmentation techniques and any LLP method that uses the proportion loss. Experimental results demonstrate this advantage and the effectiveness of our method.