ResearchTrend.AI
  • Papers
  • Communities
  • Events
  • Blog
  • Pricing
Papers
Communities
Social Events
Terms and Conditions
Pricing
Parameter LabParameter LabTwitterGitHubLinkedInBlueskyYoutube

© 2025 ResearchTrend.AI, All rights reserved.

  1. Home
  2. Papers
  3. 2403.13241
22
7

Tackling Noisy Labels with Network Parameter Additive Decomposition

20 March 2024
Jingyi Wang
Xiaobo Xia
Long Lan
Xinghao Wu
Jun-chen Yu
Wenjing Yang
Bo Han
Tongliang Liu
    NoLa
ArXivPDFHTML
Abstract

Given data with noisy labels, over-parameterized deep networks suffer overfitting mislabeled data, resulting in poor generalization. The memorization effect of deep networks shows that although the networks have the ability to memorize all noisy data, they would first memorize clean training data, and then gradually memorize mislabeled training data. A simple and effective method that exploits the memorization effect to combat noisy labels is early stopping. However, early stopping cannot distinguish the memorization of clean data and mislabeled data, resulting in the network still inevitably overfitting mislabeled data in the early training stage.In this paper, to decouple the memorization of clean data and mislabeled data, and further reduce the side effect of mislabeled data, we perform additive decomposition on network parameters. Namely, all parameters are additively decomposed into two groups, i.e., parameters w\mathbf{w}w are decomposed as w=σ+γ\mathbf{w}=\bm{\sigma}+\bm{\gamma}w=σ+γ. Afterward, the parameters σ\bm{\sigma}σ are considered to memorize clean data, while the parameters γ\bm{\gamma}γ are considered to memorize mislabeled data. Benefiting from the memorization effect, the updates of the parameters σ\bm{\sigma}σ are encouraged to fully memorize clean data in early training, and then discouraged with the increase of training epochs to reduce interference of mislabeled data. The updates of the parameters γ\bm{\gamma}γ are the opposite. In testing, only the parameters σ\bm{\sigma}σ are employed to enhance generalization. Extensive experiments on both simulated and real-world benchmarks confirm the superior performance of our method.

View on arXiv
Comments on this paper