Robust Training with Ensemble Consensus
Not on the map. Why?
Abstract (arXiv)
Since deep neural networks are over-parameterized, they can memorize noisy examples. We address such a memorization issue in the presence of label noise. From the fact that deep neural networks cannot generalize to neighborhoods of memorized features, we hypothesize that noisy examples do not consistently incur small losses on the network under a certain perturbation. Based on this, we propose a novel training method called Learning with Ensemble Consensus (LEC) that prevents overfitting to noisy examples by removing them based on the consensus of an ensemble of perturbed networks. One of the proposed LECs, LTEC outperforms the current state-of-the-art methods on noisy MNIST, CIFAR-10, and CIFAR-100 in an efficient manner.
Key facts. Cited by 11 papers in the map (data through 7 Oct 2026). First posted on arXiv 22 Oct 2019.
Builds on
No reference list yet.
Built on by
Citing papers per year: 2021: 3; 2022: 5; 2023: 1; 2024: 0; 2025: 1; 2026: 1.
11 papers.
- 2022Large Loss Matters in Weakly Supervised Multi-Label Classificationcited by 38
- 2021Learning from Noisy Labels for Entity-Centric Information Extractioncited by 34
- 2021Dual-Consistency Semi-Supervised Learning with Uncertainty Quantification for COVID-19 Lesion Segmentation from CT Imagescited by 22
- 2023ALEX: Towards Effective Graph Transfer Learning with Noisy Labelscited by 11
- 2022The Dynamic of Consensus in Deep Networks and the Identification of Noisy Labelscited by 3
- 2025Guiding Noisy Label Conditional Diffusion Models with Score-based Discriminator Correctioncited by 3
- 2022ST-CoNAL: Consistency-Based Acquisition Criterion Using Temporal Self-Ensemble for Active Learningcited by 1
- 2022Learning from Data with Noisy Labels Using Temporal Self-Ensemblecited by 1
- 2026CARE: Class-Adaptive Expert Consensus for Reliable Learning with Long-Tailed Noisy Labelscited by 1
- 2021Contrastive Representations for Label Noise Require Fine-Tuningcited by 0
- 2022Few-Example Clustering via Contrastive Learningcited by 0
- 2026CARE: Class-Adaptive Expert Consensus for Reliable Learning with Long-Tailed Noisy Labels22 May 2026
- 2025Guiding Noisy Label Conditional Diffusion Models with Score-based Discriminator Correction27 Aug 2025
- 2023ALEX: Towards Effective Graph Transfer Learning with Noisy Labels26 Sep 2023
- 2022The Dynamic of Consensus in Deep Networks and the Identification of Noisy Labels2 Oct 2022
- 2022Learning from Data with Noisy Labels Using Temporal Self-Ensemble21 Jul 2022
- 2022Few-Example Clustering via Contrastive Learning8 Jul 2022
- 2022ST-CoNAL: Consistency-Based Acquisition Criterion Using Temporal Self-Ensemble for Active Learning5 Jul 2022
- 2022Large Loss Matters in Weakly Supervised Multi-Label Classification8 Jun 2022
- 2021Contrastive Representations for Label Noise Require Fine-Tuning20 Aug 2021
- 2021Learning from Noisy Labels for Entity-Centric Information Extraction17 Apr 2021
- 2021Dual-Consistency Semi-Supervised Learning with Uncertainty Quantification for COVID-19 Lesion Segmentation from CT Images7 Apr 2021