Learning from Label Proportions with Dual-proportion Constraints

TL;DR

This paper introduces LLP-DC, a novel method for learning from label proportions that enforces dual constraints at both bag and instance levels, improving accuracy over previous methods in weakly supervised classification tasks.

Contribution

The paper proposes LLP-DC, which leverages dual proportion constraints at both levels and uses a maximum-flow algorithm for pseudo-labeling, advancing LLP techniques.

Findings

LLP-DC outperforms previous methods on benchmark datasets.

Enforcing dual constraints improves label prediction accuracy.

Method is effective across various dataset sizes and conditions.

Abstract

Learning from Label Proportions (LLP) is a weakly supervised problem in which the training data comprise bags, that is, groups of instances, each annotated only with bag-level class label proportions, and the objective is to learn a classifier that predicts instance-level labels. This setting is widely applicable when privacy constraints limit access to instance-level annotations or when fine-grained labeling is costly or impractical. In this work, we introduce a method that leverages Dual proportion Constraints (LLP-DC) during training, enforcing them at both the bag and instance levels. Specifically, the bag-level training aligns the mean prediction with the given proportion, and the instance-level training aligns hard pseudo-labels that satisfy the proportion constraint, where a minimum-cost maximum-flow algorithm is used to generate hard pseudo-labels. Extensive experimental results across various benchmark datasets empirically validate that LLP-DC consistently improves over previous LLP methods across datasets and bag sizes. The code is publicly available at https://github.com/TianhaoMa5/CV PR2026_Findings_LLP_DC.