Toward Robustness in Multi-label Classification: A Data Augmentation Strategy against Imbalance and Noise

Hwanjun Song, Minseok Kim, Jae-Gil Lee
12/12/2023
cs.LGcs.AI

Official code: DISL-Lab/BalanceMix (15 stars, MIT, last commit 2023-12-12).

Abstract

Multi-label classification poses challenges due to imbalanced and noisy labels in training data. We propose a unified data augmentation method, named BalanceMix, to address these challenges. Our approach includes two samplers for imbalanced labels, generating minority-augmented instances with high diversity. It also refines multi-labels at the label-wise granularity, categorizing noisy labels as clean, re-labeled, or ambiguous for robust optimization. Extensive experiments on three benchmark datasets demonstrate that BalanceMix outperforms existing state-of-the-art methods. We release the code at https://github.com/DISL-Lab/BalanceMix.

View on arXivView PDF

Code Implementations(1)

DISL-Lab/BalanceMixOfficial100%
MIT
150Dec 12, 20232 years agoMIT

Similar Papers

Positive Label Is All You Need for Multi-Label Classification

88% match1 repo82023

CLImage: Human-Annotated Datasets for Complementary-Label Learning

87% match1 repo172023

Free Performance Gain from Mixing Multiple Partially Labeled Samples in Multi-label Image Classification

86% match2024

Addressing Imbalance in Multi-Label Data via Label-Specific Distance-based Oversampling

86% match2026

Adaptive Gradient Calibration for Single-Positive Multi-Label Learning in Remote Sensing Image Scene Classification

85% match1 repo12025

Imprecise Label Learning: A Unified Framework for Learning with Various Imprecise Label Configurations

85% match2026

Cite this paper

@article{song2023toward,
  title  = {Toward Robustness in Multi-label Classification: A Data Augmentation Strategy against Imbalance and Noise},
  author = {Hwanjun Song and Minseok Kim and Jae-Gil Lee},
  year   = {2023},
  eprint = {2312.07087},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2312.07087v1}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2312.07087.svg)](https://codeofpaper.com/paper/2312.07087)

Discussion