Learning-with-Noisy-Labels

A curated list of most recent papers & codes in Learning with Noisy Labels

Papers & Code in 2021

This repo focus on papers after 2019, for previous works, please refer to (https://github.com/subeeshvasu/Awesome-Learning-with-Label-Noise).

ICML 2021

Conference date: Jul 18, 2021 — Jul 24, 2021

  • [UCSC REAL Lab] The importance of understanding instance-level noisy labels. [Paper]
  • [UCSC REAL Lab] Clusterability as an Alternative to Anchor Points When Learning with Noisy Labels. [Paper][Code]
  • Scaling Up Visual and Vision-Language Representation Learning With Noisy Text Supervision. [Paper][Code]
  • Learning Noise Transition Matrix from Only Noisy Labels via Total Variation Regularization. [Paper][Code]
  • Class2Simi: A Noise Reduction Perspective on Learning with Noisy Labels. [Paper]
  • Provably End-to-end Label-noise Learning without Anchor Points. [Paper]
  • Asymmetric Loss Functions for Learning with Noisy Labels. [Paper][Code]
  • Confidence Scores Make Instance-dependent Label-noise Learning Possible. [Paper]
  • Provable Generalization of SGD-trained Neural Networks of Any Width in the Presence of Adversarial Label Noise. [Paper]
  • Wasserstein Distributional Normalization For Robust Distributional Certification of Noisy Labeled Data. [Paper]
  • Learning from Noisy Labels with No Change to the Training Process. [Paper]

ICLR 2021

  • [UCSC REAL Lab] When Optimizing f-Divergence is Robust with Label Noise. [Paper][Code]
  • [UCSC REAL Lab] Learning with Instance-Dependent Label Noise: A Sample Sieve Approach. [Paper][Code]
  • Noise against noise: stochastic label noise helps combat inherent label noise. [Paper][Code]
  • Learning with Feature-Dependent Label Noise: A Progressive Approach. [Paper][Code]
  • Robust early-learning: Hindering the memorization of noisy labels. [Paper][Code]
  • MoPro: Webly Supervised Learning with Momentum Prototypes. [Paper] [Code]
  • Robust Curriculum Learning: from clean label detection to noisy label self-correction. [Paper]
  • How Does Mixup Help With Robustness and Generalization? [Paper]
  • Theoretical Analysis of Self-Training with Deep Networks on Unlabeled Data. [Paper]

CVPR 2021

Conference date: Jun 19, 2021 — Jun 25, 2021

  • [UCSC REAL Lab] A Second-Order Approach to Learning with Instance-Dependent Label Noise. [Paper][Code]
  • Improving Unsupervised Image Clustering With Robust Learning. [Paper]
  • Multi-Objective Interpolation Training for Robustness to Label Noise. [Paper][Code]
  • Noise-resistant Deep Metric Learning with Ranking-based Instance Selection. [Paper][Code]
  • Augmentation Strategies for Learning with Noisy Labels. [Paper][Code]
  • Jo-SRC: A Contrastive Approach for Combating Noisy Labels. [Paper][Code]
  • Multi-Objective Interpolation Training for Robustness to Label Noise. [Paper][Code]
  • Partially View-aligned Representation Learning with Noise-robust Contrastive Loss. [Paper][Code]
  • Correlated Input-Dependent Label Noise in Large-Scale Image Classification. [Paper]
  • DAT: Training Deep Networks Robust To Label-Noise by Matching the Feature Distributions.[Paper]
  • Faster Meta Update Strategy for Noise-Robust Deep Learning. [Paper][Code]
  • DualGraph: A graph-based method for reasoning about label noise. [Paper]
  • Background-Aware Pooling and Noise-Aware Loss for Weakly-Supervised Semantic Segmentation. [Paper]
  • Joint Negative and Positive Learning for Noisy Labels. [Paper]
  • Faster Meta Update Strategy for Noise-Robust Deep Learning. [Paper]
  • AutoDO: Robust AutoAugment for Biased Data with Label Noise via Scalable Probabilistic Implicit Differentiation. [Paper][Code]
  • Meta Pseudo Labels. [Paper][Code]
  • All Labels Are Not Created Equal: Enhancing Semi-supervision via Label Grouping and Co-training. [Paper][Code]
  • SimPLE: Similar Pseudo Label Exploitation for Semi-Supervised Classification. [Paper][Code]

AISTATS 2021

Conference date: Apr 13, 2021 — Apr 15, 2021

  • Collaborative Classification from Noisy Labels. [Paper]
  • Linear Models are Robust Optimal Under Strategic Behavior. [Paper]

AAAI 2021

  • Beyond Class-Conditional Assumption: A Primary Attempt to Combat Instance-Dependent Label Noise. [Paper][Code]
  • Learning to Purify Noisy Labels via Meta Soft Label Corrector. [Paper][Code]
  • Robustness of Accuracy Metric and its Inspirations in Learning with Noisy Labels. [Paper][Code]
  • Learning from Noisy Labels with Complementary Loss Functions. [Paper][Code]
  • Analysing the Noise Model Error for Realistic Noisy Label Data. [Paper][Code]
  • Tackling Instance-Dependent Label Noise via a Universal Probabilistic Model. [Paper]
  • Learning with Group Noise. [Paper]
  • Meta Label Correction for Noisy Label Learning. [Paper]

ArXiv 2021

  • [UCSC REAL Lab] Understanding (Generalized) Label Smoothing when Learning with Noisy Labels. [Paper]
  • Pervasive Label Errors in Test Sets Destabilize Machine Learning Benchmarks. [Paper][Code]
  • Estimating Instance-dependent Label-noise Transition Matrix using DNNs. [Paper]
  • A Theoretical Analysis of Learning with Noisily Labeled Data. [Paper]
  • Generalized Jensen-Shannon Divergence Loss for Learning with Noisy Labels. [Paper]
  • A Survey of Label-noise Representation Learning: Past, Present and Future. [Paper]
  • Learning Noise Transition Matrix from Only Noisy Labels via Total Variation Regularization. [Paper][Code]
  • Noisy-Labeled NER with Confidence Estimation. [Paper][Code]
  • Study Group Learning: Improving Retinal Vessel Segmentation Trained with Noisy Labels. [Paper][Code]
  • Contrast to Divide: Self-Supervised Pre-Training for Learning with Noisy Labels. [Paper][Code]
  • Exponentiated Gradient Reweighting for Robust Training Under Label Noise and Beyond. [Paper]
  • Understanding the Interaction of Adversarial Training with Noisy Labels. [Paper]
  • Learning from Noisy Labels via Dynamic Loss Thresholding. [Paper]
  • Evaluating Multi-label Classifiers with Noisy Labels. [Paper]
  • Self-Supervised Noisy Label Learning for Source-Free Unsupervised Domain Adaptation. [Paper]
  • Transform consistency for learning with noisy labels. [Paper]
  • Learning to Combat Noisy Labels via Classification Margins. [Paper]
  • Joint Negative and Positive Learning for Noisy Labels. [Paper]
  • Robust Classification from Noisy Labels: Integrating Additional Knowledge for Chest Radiography Abnormality Assessment. [Paper]
  • DST: Data Selection and joint Training for Learning with Noisy Labels. [Paper]
  • LongReMix: Robust Learning with High Confidence Samples in a Noisy Label Environment. [Paper]
  • A Novel Perspective for Positive-Unlabeled Learning via Noisy Labels. [Paper]
  • Ensemble Learning with Manifold-Based Data Splitting for Noisy Label Correction. [Paper]
  • MetaLabelNet: Learning to Generate Soft-Labels from Noisy-Labels. [Paper]
  • On the Robustness of Monte Carlo Dropout Trained with Noisy Labels. [Paper]
  • Co-matching: Combating Noisy Labels by Augmentation Anchoring. [Paper]
  • Pathological Image Segmentation with Noisy Labels. [Paper]
  • CrowdTeacher: Robust Co-teaching with Noisy Answers & Sample-specific Perturbations for Tabular Data. [Paper]
  • Approximating Instance-Dependent Noise via Instance-Confidence Embedding. [Paper]
  • Rethinking Noisy Label Models: Labeler-Dependent Noise with Adversarial Awareness. [Paper]
  • ScanMix: Learning from Severe Label Noise viaSemantic Clustering and Semi-Supervised Learning. [Paper]
  • Friends and Foes in Learning from Noisy Labels. [Paper]
  • Learning from Noisy Labels for Entity-Centric Information Extraction. [Paper]
  • A Fremework Using Contrastive Learning for Classification with Noisy Labels. [Paper]
  • Contrastive Learning Improves Model Robustness Under Label Noise. [Paper][Code]
  • Noise-Resistant Deep Metric Learning with Probabilistic Instance Filtering. [Paper]
  • Compensation Learning. [Paper]
  • kNet: A Deep kNN Network To Handle Label Noise. [Paper]
  • Temporal-aware Language Representation Learning From Crowdsourced Labels. [Paper]
  • Memorization in Deep Neural Networks: Does the Loss Function matter?. [Paper]
  • Mitigating Memorization in Sample Selection for Learning with Noisy Labels. [Paper]
  • P-DIFF: Learning Classifier with Noisy Labels based on Probability Difference Distributions. [Paper][Code]
  • Decoupling Representation and Classifier for Noisy Label Learning. [Paper]
  • Contrastive Representations for Label Noise Require Fine-Tuning. [Paper]
  • NGC: A Unified Framework for Learning with Open-World Noisy Data. [Paper]
  • Learning From Long-Tailed Data With Noisy Labels. [Paper]
  • Robust Long-Tailed Learning Under Label Noise. [Paper]
  • Instance-dependent Label-noise Learning under a Structural Causal Model. [Paper]
  • Assessing the Quality of the Datasets by Identifying Mislabeled Samples. [Paper]
  • Learning to Aggregate and Refine Noisy Labels for Visual Sentiment Analysis. [Paper]
  • Assessing the Quality of the Datasets by Identifying Mislabeled Samples. [Paper]

Papers & Code in 2020

ICML 2020

  • [UCSC REAL Lab] Peer Loss Functions: Learning from Noisy Labels without Knowing Noise Rates. [Paper][Code 1] [Code 2]
  • Normalized Loss Functions for Deep Learning with Noisy Labels. [Paper][Code]
  • SIGUA: Forgetting May Make Learning with Noisy Labels More Robust. [Paper][Code]
  • Error-Bounded Correction of Noisy Labels. [Paper][Code]
  • Training Binary Neural Networks through Learning with Noisy Supervision. [Paper][Code]
  • Improving generalization by controlling label-noise information in neural network weights. [Paper][Code]
  • Self-PU: Self Boosted and Calibrated Positive-Unlabeled Training. [Paper][Code]
  • Searching to Exploit Memorization Effect in Learning with Noisy Labels. [Paper][Code]
  • Learning with Bounded Instance and Label-dependent Label Noise. [Paper]
  • Label-Noise Robust Domain Adaptation. [Paper]
  • Beyond Synthetic Noise: Deep Learning on Controlled Noisy Labels. [Paper]
  • Does label smoothing mitigate label noise?. [Paper]
  • Learning with Multiple Complementary Labels. [Paper]
  • Deep k-NN for Noisy Labels. [Paper]
  • Extreme Multi-label Classification from Aggregated Labels. [Paper]

ICLR 2020

  • DivideMix: Learning with Noisy Labels as Semi-supervised Learning. [Paper][Code]
  • Learning from Rules Generalizing Labeled Exemplars. [Paper] [Code]
  • Robust training with ensemble consensus. [Paper][Code]
  • Self-labelling via simultaneous clustering and representation learning. [Paper][Code]
  • Can gradient clipping mitigate label noise? [Paper][Code]
  • Mutual Mean-Teaching: Pseudo Label Refinery for Unsupervised Domain Adaptation on Person Re-identification. [Paper][Code]
  • Curriculum Loss: Robust Learning and Generalization against Label Corruption. [Paper]
  • Simple and Effective Regularization Methods for Training on Noisily Labeled Data with Generalization Guarantee. [Paper]
  • SELF: Learning to Filter Noisy Labels with Self-Ensembling. [Paper]

Nips 2020

  • Part-dependent Label Noise: Towards Instance-dependent Label Noise. [Paper][Code]
  • Identifying Mislabeled Data using the Area Under the Margin Ranking. [Paper][Code]
  • Dual T: Reducing Estimation Error for Transition Matrix in Label-noise Learning. [Paper]
  • Early-Learning Regularization Prevents Memorization of Noisy Labels. [Paper][Code]
  • Coresets for Robust Training of Deep Neural Networks against Noisy Labels. [Paper][Code]
  • Modeling Noisy Annotations for Crowd Counting. [Paper][Code]
  • Robust Optimization for Fairness with Noisy Protected Groups. [Paper][Code]
  • Stochastic Optimization with Heavy-Tailed Noise via Accelerated Gradient Clipping. [Paper][Code]
  • A Topological Filter for Learning with Label Noise. [Paper][Code]
  • Self-Adaptive Training: beyond Empirical Risk Minimization. [Paper][Code]
  • Disentangling Human Error from the Ground Truth in Segmentation of Medical Images. [Paper][Code]
  • Non-Convex SGD Learns Halfspaces with Adversarial Label Noise. [Paper]
  • Efficient active learning of sparse halfspaces with arbitrary bounded noise. [Paper]
  • Semi-Supervised Partial Label Learning via Confidence-Rated Margin Maximization. [Paper]
  • Labelling unlabelled videos from scratch with multi-modal self-supervision. [Paper][Code]
  • Distribution Aligning Refinery of Pseudo-label for Imbalanced Semi-supervised Learning. [Paper][Code]
  • MetaPoison: Practical General-purpose Clean-label Data Poisoning. [Paper][Code 1][Code 2]
  • Provably Consistent Partial-Label Learning. [Paper]
  • A Variational Approach for Learning from Positive and Unlabeled Data. [Paper][Code]

AAAI 2020

  • [UCSC REAL Lab] Reinforcement Learning with Perturbed Rewards. [Paper] [Code]
  • Less Is Better: Unweighted Data Subsampling via Influence Function. [Paper] [Code]
  • Weakly Supervised Sequence Tagging from Noisy Rules. [Paper][Code]
  • Coupled-View Deep Classifier Learning from Multiple Noisy Annotators. [Paper]
  • Partial multi-label learning with noisy label identification. [Paper]
  • Self-Paced Robust Learning for Leveraging Clean Labels in Noisy Data. [Paper]
  • Label Error Correction and Generation Through Label Relationships. [Paper]

CVPR 2020

  • Combating noisy labels by agreement: A joint training method with co-regularization. [Paper][Code]
  • Distilling Effective Supervision From Severe Label Noise. [Paper][Code]
  • Self-Training With Noisy Student Improves ImageNet Classification. [Paper][Code]
  • Noise Robust Generative Adversarial Networks. [Paper][Code]
  • Global-Local GCN: Large-Scale Label Noise Cleansing for Face Recognition. [Paper]
  • DLWL: Improving Detection for Lowshot Classes With Weakly Labelled Data. [Paper]
  • Spherical Space Domain Adaptation With Robust Pseudo-Label Loss. [Paper][Code]
  • Training Noise-Robust Deep Neural Networks via Meta-Learning. [Paper][Code]
  • Shoestring: Graph-Based Semi-Supervised Classification With Severely Limited Labeled Data. [Paper][Code]
  • Noise-Aware Fully Webly Supervised Object Detection. [Paper][Code]
  • Learning From Noisy Anchors for One-Stage Object Detection. [Paper][Code]
  • Generating Accurate Pseudo-Labels in Semi-Supervised Learning and Avoiding Overconfident Predictions via Hermite Polynomial Activations. [Paper][Code]
  • Revisiting Knowledge Distillation via Label Smoothing Regularization. [Paper][Code]

ECCV 2020

  • 2020-ECCV – Learning with Noisy Class Labels for Instance Segmentation. [Paper][Code]
  • 2020-ECCV – Suppressing Mislabeled Data via Grouping and Self-Attention. [Paper][Code]
  • 2020-ECCV – NoiseRank: Unsupervised Label Noise Reduction with Dependence Models. [Paper]
  • 2020-ECCV – Weakly Supervised Learning with Side Information for Noisy Labeled Images. [Paper]
  • 2020-ECCV – Learning Noise-Aware Encoder-Decoder from Noisy Labels by Alternating Back-Propagation for Saliency Detection. [Paper]
  • 2020-ECCV – Graph convolutional networks for learning with few clean and many noisy labels. [Paper]

ArXiv 2020

  • No Regret Sample Selection with Noisy Labels. [Paper][Code]
  • Meta Soft Label Generation for Noisy Labels. [Paper][Code]
  • Learning from Noisy Labels with Deep Neural Networks: A Survey. [Paper]
  • RAR-U-Net: a Residual Encoder to Attention Decoder by Residual Connections Framework for Spine Segmentation under Noisy Labels.
  • Learning from Small Amount of Medical Data with Noisy Labels: A Meta-Learning Approach. [Paper]

GitHub

https://github.com/weijiaheng/Advances-in-Label-Noise-Learning