Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Controller-Guided Partial Label Consistency Regularization with Unlabeled Data

Qian-Wei Wang, Bowen Zhao, Mingyan Zhu, Tianxiang Li, Zimo Liu, Shu-Tao Xia

TL;DR

Partial label learning (PLL) often struggles when supervision is sparse. ConCont introduces a controller-guided consistency framework that leverages unlabeled data to regularize both label-level and representation-level predictions, guided by a per-example p-score and adaptive class-balanced thresholds. The method combines a partial-label loss with controller-driven consistency losses and per-class threshold updates, demonstrating substantial gains over state-of-the-art PLL baselines across CIFAR-10/100 and SVHN under limited annotation. Its modular design enables integration with existing PLL approaches, broadening practical PLL deployment where annotation resources are constrained.

Abstract

Partial label learning (PLL) learns from training examples each associated with multiple candidate labels, among which only one is valid. In recent years, benefiting from the strong capability of dealing with ambiguous supervision and the impetus of modern data augmentation methods, consistency regularization-based PLL methods have achieved a series of successes and become mainstream. However, as the partial annotation becomes insufficient, their performances drop significantly. In this paper, we leverage easily accessible unlabeled examples to facilitate the partial label consistency regularization. In addition to a partial supervised loss, our method performs a controller-guided consistency regularization at both the label-level and representation-level with the help of unlabeled data. To minimize the disadvantages of insufficient capabilities of the initial supervised model, we use the controller to estimate the confidence of each current prediction to guide the subsequent consistency regularization. Furthermore, we dynamically adjust the confidence thresholds so that the number of samples of each class participating in consistency regularization remains roughly equal to alleviate the problem of class-imbalance. Experiments show that our method achieves satisfactory performances in more practical situations, and its modules can be applied to existing PLL methods to enhance their capabilities.

Controller-Guided Partial Label Consistency Regularization with Unlabeled Data

TL;DR

Partial label learning (PLL) often struggles when supervision is sparse. ConCont introduces a controller-guided consistency framework that leverages unlabeled data to regularize both label-level and representation-level predictions, guided by a per-example p-score and adaptive class-balanced thresholds. The method combines a partial-label loss with controller-driven consistency losses and per-class threshold updates, demonstrating substantial gains over state-of-the-art PLL baselines across CIFAR-10/100 and SVHN under limited annotation. Its modular design enables integration with existing PLL approaches, broadening practical PLL deployment where annotation resources are constrained.

Abstract

Partial label learning (PLL) learns from training examples each associated with multiple candidate labels, among which only one is valid. In recent years, benefiting from the strong capability of dealing with ambiguous supervision and the impetus of modern data augmentation methods, consistency regularization-based PLL methods have achieved a series of successes and become mainstream. However, as the partial annotation becomes insufficient, their performances drop significantly. In this paper, we leverage easily accessible unlabeled examples to facilitate the partial label consistency regularization. In addition to a partial supervised loss, our method performs a controller-guided consistency regularization at both the label-level and representation-level with the help of unlabeled data. To minimize the disadvantages of insufficient capabilities of the initial supervised model, we use the controller to estimate the confidence of each current prediction to guide the subsequent consistency regularization. Furthermore, we dynamically adjust the confidence thresholds so that the number of samples of each class participating in consistency regularization remains roughly equal to alleviate the problem of class-imbalance. Experiments show that our method achieves satisfactory performances in more practical situations, and its modules can be applied to existing PLL methods to enhance their capabilities.
Paper Structure (23 sections, 10 equations, 5 figures, 6 tables)

This paper contains 23 sections, 10 equations, 5 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Methodology
  4. Notations
  5. Consistency Regularization with Controller
  6. Adaptive Confidence Threshold
  7. Experiments
  8. Experimental Setup
  9. Main Results
  10. Ablation Studies
  11. Effects of supervised loss and consistency regularization.
  12. Comparison of different class-imbalance methods.
  13. Different selection strategies of positive and negative samples.
  14. Conclusion
  15. Acknowledgments
  16. ...and 8 more sections

Figures (5)

  • Figure 1: The performances of ConCont (with additional unlabeled examples), DPLL and PiCO with $100\%$, $50\%$, $20\%$ and $10\%$ partially annotated examples on CIFAR-10 dataset.
  • Figure 2: Partial cross-entropy loss transforms multi-class predicted distribution into binary.
  • Figure 3: The overall framework of ConCont. Dotted lines represent sharing backbone.
  • Figure 4: The P-R curves of classifying confident or unconfident examples using p-scores and maximum probabilities, experiments conducted on CIFAR-100 ($40\%$).
  • Figure 5: Momentum constrast.