Abstract:Learning from label proportions (LLP) is a kind of weakly supervised learning that trains an instance-level classifier from label proportions of bags, which consist of sets of instances without using instance labels. A challenge in LLP arises when the number of instances in a bag (bag size) is numerous, making the traditional LLP methods difficult due to GPU memory limitations. This study aims to develop an LLP method capable of learning from bags with large sizes. In our method, smaller bags (mini-bags) are generated by sampling instances from large-sized bags (original bags), and these mini-bags are used in place of the original bags. However, the proportion of a mini-bag is unknown and differs from that of the original bag, leading to overfitting. To address this issue, we propose a perturbation method for the proportion labels of sampled mini-bags to mitigate overfitting to noisy label proportions. This perturbation is added based on the multivariate hypergeometric distribution, which is statistically modeled. Additionally, loss weighting is implemented to reduce the negative impact of proportions sampled from the tail of the distribution. Experimental results demonstrate that the proportion label perturbation and loss weighting achieve classification accuracy comparable to that obtained without sampling. Our codes are available at <a class="link-external link-https" href="https://github.com/stainlessnight/LLP-LargeBags" rel="external noopener nofollow">this https URL</a>.

$\propto$SVM for learning with label proportions

A Novel Svm Modeling Approach For Highly Imbalanced And Overlapping Classification

A Multi-Class Large Margin Classifier

Semi-Supervised Learning with Label Proportion

Two-stage Training for Learning from Label Proportions

OT-LLP: Optimal Transport for Learning from Label Proportions

Robust Multi-Label Learning with PRO Loss

Probabilistic Labeled Semi-supervised SVM

Learning a Generative Classifier from Label Proportions

Nonparametric semi-supervised learning of class proportions

Theoretical Proportion Label Perturbation for Learning from Label Proportions in Large Bags

Learning from Label Proportions with Instance-wise Consistency

Learning from Label Proportions by Learning with Label Noise

Learning from Label Proportions: Bootstrapping Supervised Learners via Belief Propagation

Learning from Label Proportions and Covariate-shifted Instances

Safe instance screening for primal multi-label ProSVM

Approximate Approach to Train SVM on Very Large Data Sets

Transfer Learning-Based Label Proportions Method with Data of Uncertainty

Learning from Label Proportions with Consistency Regularization

Learning from Label Proportions with Generative Adversarial Networks

SVM Algorithms for Large Scale Classification Problems Based on Data Partition and Ensemble Learning