aireadinggroup / noise-label

论文收集

DNN噪声标签相关论文

综述

• Learning from Noisy Labels with Deep Neural Networks: A Survey

类型	方法
Robust Loss Function	Robust MAE, Generalized Cross Entropy,Symmetric Cross Entropy, Curriculum Loss
Robust Architecture	Webly Learning, Noise Model, Dropout Noise Model, S–model , C–model , NLNN , Probabilistic Noise Model , Masking, Contrastive-Additive Noise Network
Robust Regularization	Adversarial Training , Label Smoothing , Mixup , Bilevel Learning , Annotator Confusion , Pre-training
Loss Adjustment	Gold Loss Correction, Importance Reweighting, Active Bias , Bootstrapping, Dynamic Bootstrapping , D2L , SELFIE
Sample Selection	Decouple , MentorNet , Co-teaching , Co-teaching+, Iterative Detection , ITLM , INCV , SELFIE , SELF , Curriculum Loss
Meta Learning	Meta-Regressor , Knowledge Distillation , L2LWS , CWS , Automatic Reweighting , MLNT , Meta-Weight-Net, Data Coefficients
Semi-supervised Learning	Label Aggregation , Two-Stage Framework , SELF , DivideMix

Robust Loss Function

• (Robust MAE) Robust Loss Functions under Label Noise for Deep Neural Networks pdf
• (Generalized Cross Entropy) Generalized cross entropy loss for training deep neural networks with noisy labels pdf
• (Symmetric Cross Entropy )Symmetric cross entropy for robust learning with noisy labels pdf
• (Curriculum Loss ) Curriculum loss: Robust learning and generalization against label corruptionpdf

Robust Architecture

• (Webly Learning) Webly supervised learning of convolutional networks pdf
• (Noise Model)Training convolutional networks with noisy labels pdf
• (Dropout Noise Model)Learning deep networks from noisy labels with dropout regularization pdf
• (S–model) (C–model)Training deep neural-networks using a noise adaptation layer pdf
• (NLNN)Training deep neural-networks based on unreliable labels pdf
• (Probabilistic Noise Model) Learning from massive noisy labeled data for image classification pdf
• (Masking) Masking: A new perspective of noisy supervision pdf
• (Contrastive-Additive Noise Network) Deep learning from noisy image labels with quality embedding pdf

Robust Regularization

• (Adversarial Training) Explaining and harnessing adversarial examples pdf
• (Label Smoothing) Regularizing neural networks by penalizing confident output distributions pdf
• (Mixup)mixup: Beyond empirical risk minimization pdf
• (Bilevel Learning)Deep bilevel learningpdf
• (Annotator Confusion) Learning from noisy labels by regularized estimation of annotator confusion pdf
• (Pre-training) Using Pre-Training Can Improve Model Robustness and Uncertainty pdf

Loss Adjustment

• (Gold Loss Correction) Using trusted data to train deep networks on labels corrupted by severe noisepdf
• (Importance Reweighting) Multiclass learning with partially corrupted labels pdf
• (Active Bias) Active Bias: Training more accurate neural networks by emphasizing high variance samples pdf
• (Bootstrapping) Training deep neural networks on noisy labels with bootstrapping pdf
• (Dynamic Bootstrapping) Unsupervised label noise modeling and loss correction pdf
• (D2L) Dimensionality-driven learning with noisy labels pdf
• (SELFIE ) SELFIE: Refurbishing unclean samples for robust deep learning pdf

Sample Selection

• Decouple “Decoupling” when to update” from” how to update” pdf
• MentorNet MentorNet: Learning data-driven curriculum for very deep neural networks on corrupted label pdf
• Co-teaching Co-teaching: Robust training of deep neural networks with extremely noisy labels pdf
• Co-teaching+ How does disagreement help generalization against label corruption? pdf
• Iterative Detection Iterative learning with open-set noisy labels pdf
• ITLM Iterative learning with open-set noisy labels pdf
• INCV Understanding and utilizing deep neural networks trained with noisy labels pdf
• SELFIE SELFIE: Refurbishing unclean samples for robust deep learning pdf
• SELF Self: Learning to filter noisy labels with self-ensembling pdf
• Curriculum Loss Curriculum loss: Robust learning and generalization against label corruption pdf

Meta Learning

• Meta-Regressor Noise detection in the meta-learning level pdf
• Knowledge Distillation Learning from noisy labels with distillation pdf
• L2LWS Learning to Learn from Weak Supervision by Full Supervision pdf
• CWS Avoiding your teacher’s mistakes: Training neural networks with controlled weak supervision pdf
• Automatic Reweighting Learning to reweight examples for robust deep learning pdf
• MLNT Learning to learn from noisy labeled data pdf
• Meta-Weight-Net MetaWeight-Net: Learning an explicit mapping for sample weighting pdf
• Data Coefficients MetaWeight-Net: Learning an explicit mapping for sample weighting pdf

Semi-supervised Learning

• Label Aggregation Robust semisupervised learning through label aggregation pdf
• Two-Stage Framework A semi-supervised two-stage approach to learning from noisy labels pdf
• SELF Self: Learning to filter noisy labels with self-ensembling pdf
• DivideMix DivideMix: Learning with noisy labels as semi-supervised learning pdf

其他

• Robust deep supervised hashing for image retrieval

ICLR 2021

• NOISE-ROBUST CONTRASTIVE LEARNING pdf
• ROBUST CURRICULUM LEARNING: FROM CLEAN LABEL DETECTION TO NOISY LABEL SELF-CORRECTION pdf
• ROBUST LEARNING VIA GOLDEN SYMMETRIC LOSS OF (UN)TRUSTED LABELS pdf
• LONG-TAIL ZERO AND FEW-SHOT LEARNING VIA CONTRASTIVE PRETRAINING ON AND FOR SMALL DATA pdf
• A SIMPLE APPROACH TO DEFINE CURRICULA FOR TRAINING NEURAL NETWORKS pdf
• INTERACTIVE WEAK SUPERVISION: LEARNING USEFUL HEURISTICS FOR DATA LABELING pdf
• PROTOTYPICAL REPRESENTATION LEARNING FOR RELATION EXTRACTION pdf
• NEIGHBOR CLASS CONSISTENCY ON UNSUPERVISEDDOMAIN ADAPTATION pdf
• ON THE IMPORTANCE OF LOOKING AT THE MANIFOLD pdf
• SHAPE OR TEXTURE: UNDERSTANDING DISCRIMINATIVE FEATURES IN CNNS pdf
• LEARNING TO EXPLORE WITH PLEASURE pdf