Leveraging Label Proportion Prior for Class-Imbalanced Semi-Supervised Learning

Kohki Akiba; Shinnosuke Matsuo; Shota Harada; Ryoma Bise

cs.LGMar 3, 2026

Leveraging Label Proportion Prior for Class-Imbalanced Semi-Supervised Learning

Authors: Kohki Akiba, Shinnosuke Matsuo, Shota Harada, Ryoma Bise

Organizations: Kyushu University, Fukuoka, Japan

Abstract

Semi-supervised learning (SSL) often suffers under class imbalance, where pseudo-labeling amplifies majority bias and suppresses minority performance. We address this issue with a lightweight framework that, to our knowledge, is the first to introduce Proportion Loss from learning from label proportions (LLP) into SSL as a regularization term. Proportion Loss aligns model predictions with the global class distribution, mitigating bias across both majority and minority classes. To further stabilize training, we formulate a stochastic variant that accounts for fluctuations in mini-batch composition. Experiments on the Long-tailed CIFAR-10 benchmark show that integrating Proportion Loss into FixMatch and ReMixMatch consistently improves performance over the baselines across imbalance severities and label ratios, and achieves competitive or superior results compared to existing CISSL methods, particularly under scarce-label conditions.

Figures & tables

Figure 1: Our approach: Regularization with the label proportion prior for semi-supervised learning (SSL).

Figure 2: Overview of the proposed method.

Method	$\gamma=10$ , $\beta=2\%$	$\gamma=20$ , $\beta=4\%$	$\gamma=50$ , $\beta=10\%$	$\gamma=100$ , $\beta=20\%$
FixMatch	80.8±0.11	78.2±0.14	74.5±0.19	71.1±0.22
FixMatch + DARP	81.3±0.10	79.1±0.13	74.3±0.19	71.0±0.23
FixMatch + (CReST+PDA)	79.6±0.12	76.4±0.14	76.6±0.16	73.0±0.19
FixMatch + Ours	81.9±0.09	80.4±0.11	75.7±0.17	72.8±0.20
ReMixMatch	85.5±0.08	83.4±0.10	79.3±0.14	75.5±0.19
ReMixMatch + DARP	87.5±0.08	85.3±0.09	81.5±0.13	77.8±0.17

Table 1: Test accuracy (%) on CIFAR-10-LT under different imbalance ratios

\gamma

and labeled data ratios

\beta

. Results are averaged over five runs with different random seeds.

Panel of Figure 3 — Figure 3: Comparison of estimated output proportions after training. Red indicates overestimation and blue indicates underestimation across classes (Class 1 = major, Classes 7–9 = minor).

Panel of Figure 4 — Figure 4: Recall of pseudo-labels for the most major class (left) and most minor class (right) during training.

Semi-Supervised Learning

Semi-supervised learning uses both labeled and unlabeled data to improve prediction when reliable annotations are scarce.

See all 164 papers

Geometric Regularization for Long-Tailed Semi-Supervised Learning via Gaussian Feature Bridges
Aug 21, 2026Hongyang He, Xinyuan Song, Yan Zhong +4Semi-Supervised Learning Long-Tailed Classification
University of Warwick, Coventry, United Kingdom · Wuhan University, Wuhan, China · Emory University, Atlanta, GA, USA +2
Learning Dynamics of Logits Debiasing for Long-Tailed Semi-Supervised Learning
Aug 31, 2026Yue Cheng, Jiajun Zhang, Xiaohui Gao +2Class-Imbalanced Learning Semi-Supervised Learning
Beijing Jiaotong University · AntGroup · Northwestern Polytechnical University +1
Learning from Label Proportions with Dual-proportion Constraints
Mar 22, 2026Tianhao Ma, Ximing Li, Changchun Li +1Weakly Supervised Learning
The University of Tokyo, Japan · College of Computer Science and Technology, Jilin University, China · Key Laboratory of Symbolic Computation and Knowledge Engineering of Ministry of Education, Jilin University, China +1

Aug 21, 2026J/K move · Enter open · S save

Geometric Regularization for Long-Tailed Semi-Supervised Learning via Gaussian Feature Bridges

Hongyang He, Xinyuan Song, Yan Zhong, Daizong Liu+3

University of Warwick, Coventry, United Kingdom · Wuhan University, Wuhan, China · Emory University, Atlanta, GA, USA+2

Semi-Supervised Learning Long-Tailed Classification Prototype-Based Learning

Open paper AskPDF arXiv

Real-world semi-supervised learning (SSL) often encounters significant challenges with long-tailed label distributions and noisy pseudo-labels, which hinder generalization and amplify confirmation bias. In this work, we introduce a novel framework, Gaussian Bridge Consistency (GBC), to address these challenges by constructing semantic interpolation paths between unlabeled samples and high-quality class anchors. Our method maintains a dynamic Prototype Atlas that stores a diverse and evolving set of labeled and pseudo-labeled exemplars per class. For each unlabeled instance, GBC forms a class-conditional Gaussian Feature Bridge in the latent space, enabling the student model to traverse a smooth trajectory from uncertain predictions to reliable class prototypes. A bridge consistency loss is applied along this path to enforce alignment with a geometrically interpolated target distribution. Furthermore, we propose BridgeMix, a confidence-aware feature mixing strategy that interpolates both sample and anchor pairs to amplify cross-sample generalization. Extensive experiments on CIFAR10-LT and ImageNet-LT (USB benchmarks) validate the robustness and effectiveness of GBC under realistic long-tailed SSL settings, consistently improving long tail-class performance without sacrificing scalability.

Abstract

Semi-Supervised Learning

Explore similar work

Geometric Regularization for Long-Tailed Semi-Supervised Learning via Gaussian Feature Bridges

Learning Dynamics of Logits Debiasing for Long-Tailed Semi-Supervised Learning

Learning from Label Proportions with Dual-proportion Constraints