NeurIPS 2021 accepted papers on Label-efficient and robust learning
262 of the 2,508 papers accepted at NeurIPS 2021 (10.4%) are on Label-efficient and robust learning, compared with 9.2% at NeurIPS 2020. They fall into 9 topics.
See NeurIPS 2021 in the interactive map
Submitting to NeurIPS? Checklist →
Working on something in this area? Describe your idea in scime atlas to see the closest papers by meaning and how crowded the spot has become.
Topics and papers
Recognition and long-tail · 102 papers
- TransMIL: Transformer based Correlated Multiple Instance Learning for Whole Slide Image Classification (505 citations)
- Relative Uncertainty Learning for Facial Expression Recognition (77 citations)
- Learning Debiased Representation via Disentangled Feature Augmentation (46 citations)
- WaveFake: A Data Set to Facilitate Audio Deepfake Detection (44 citations)
- Rethinking Calibration of Deep Neural Networks: Do Not Be Afraid of Overconfidence (39 citations)
- COVID-19 Sounds: A Large-Scale Audio Dataset for Digital Respiratory Screening (38 citations)
- WRENCH: A Comprehensive Benchmark for Weak Supervision (38 citations)
- BiToD: A Bilingual Multi-Domain Dataset For Task-Oriented Dialogue Modeling (35 citations)
- Curriculum Disentangled Recommendation with Noisy Multi-feedback (31 citations)
- AutoBalance: Optimized Loss Functions for Imbalanced Data (28 citations)
- Fairness via Representation Neutralization (27 citations)
- Automatic Unsupervised Outlier Model Selection (26 citations)
and 90 more in the interactive map.
Self-supervised and contrastive · 36 papers
- Partial success in closing the gap between human and machine vision (63 citations)
- Your head is there to move you around: Goal-driven models of the primate dorsal pathway (54 citations)
- From Canonical Correlation Analysis to Self-supervised Graph Neural Networks (49 citations)
- Predicting What You Already Know Helps: Provable Self-Supervised Learning (46 citations)
- Provable Guarantees for Self-Supervised Deep Learning with Spectral Contrastive Loss (41 citations)
- ReSSL: Relational Self-Supervised Learning with Weak Augmentation (41 citations)
- The functional specialization of visual cortex emerges from training parallel pathways with self-supervised predictive learning (38 citations)
- Self-Supervised Bug Detection and Repair (36 citations)
- Do Different Tracking Tasks Require Different Appearance Models? (32 citations)
- Unleashing the Power of Contrastive Self-Supervised Visual Models via Contrast-Regularized Fine-Tuning (31 citations)
- Self-Supervised Learning with Kernel Dependence Maximization (21 citations)
- Unsupervised Speech Recognition (21 citations)
and 24 more in the interactive map.
Domain adaptation and generalization · 34 papers
- Exploiting the Intrinsic Neighborhood Structure for Source-free Domain Adaptation (103 citations)
- Model Adaptation: Historical Contrastive Learning for Unsupervised Domain Adaptation without Source Data (95 citations)
- Exploiting Domain-Specific Features to Enhance Domain Generalization (70 citations)
- Cycle Self-Training for Domain Adaptation (61 citations)
- Leveraging Distribution Alignment via Stein Path for Cross-Domain Cold-Start Recommendation (38 citations)
- Confident Anchor-Induced Multi-Source Free Domain Adaptation (37 citations)
- A Prototype-Oriented Framework for Unsupervised Domain Adaptation (24 citations)
- Gradual Domain Adaptation without Indexed Intermediate Domains (13 citations)
- Adversarial Reweighting for Partial Domain Adaptation (12 citations)
- CLDA: Contrastive Learning for Semi-Supervised Domain Adaptation (12 citations)
- A Mathematical Framework for Quantifying Transferability in Multi-source Transfer Learning (11 citations)
- An Empirical Investigation of Domain Generalization with Empirical Risk Minimizers (10 citations)
and 22 more in the interactive map.
Zero-/few-shot and meta-learning · 23 papers
- HSVA: Hierarchical Semantic-Visual Adaptation for Zero-Shot Learning (85 citations)
- Rectifying the Shortcut Learning of Background for Few-Shot Learning (45 citations)
- Independent Prototype Propagation for Zero-Shot Compositionality (23 citations)
- Dual Progressive Prototype Network for Generalized Zero-Shot Learning (22 citations)
- Re-ranking for image retrieval and transductive few-shot classification (21 citations)
- Fine-Grained Zero-Shot Learning with DNA as Side Information (19 citations)
- POODLE: Improving Few-shot Learning via Penalizing Out-of-Distribution Samples (19 citations)
- On Episodes, Prototypical Networks, and Few-Shot Learning (15 citations)
- FS-Mol: A Few-Shot Learning Dataset of Molecules (11 citations)
- Dynamic Distillation Network for Cross-Domain Few-Shot Recognition with Unlabeled Data (10 citations)
- Grad2Task: Improved Few-shot Text Classification Using Gradients for Task Representation (10 citations)
- Weak-shot Fine-grained Classification via Similarity Transfer (9 citations)
and 11 more in the interactive map.
OOD and anomaly detection · 18 papers
- Exploring the Limits of Out-of-Distribution Detection (103 citations)
- Learning Causal Semantic Representation for Out-of-Distribution Prediction (47 citations)
- ReAct: Out-of-distribution Detection With Rectified Activations (46 citations)
- Invariance Principle Meets Information Bottleneck for Out-of-Distribution Generalization (44 citations)
- Towards a Theoretical Framework of Out-of-Distribution Generalization (39 citations)
- On the Importance of Gradients for Detecting Distributional Shifts in the Wild (38 citations)
- Can multi-label classification networks know what they don’t know? (35 citations)
- STEP: Out-of-Distribution Detection in the Presence of Limited In-Distribution Labeled Data (10 citations)
- Reliable and Trustworthy Machine Learning for Health Using Dataset Shift Detection (7 citations)
- Detecting Errors and Estimating Accuracy on Unlabeled Data with Self-training Ensembles (6 citations)
- On the Out-of-distribution Generalization of Probabilistic Image Modelling (3 citations)
- Single Layer Predictive Normalized Maximum Likelihood for Out-of-Distribution Detection (3 citations)
and 6 more in the interactive map.
Semi-supervised learning · 14 papers
- FlexMatch: Boosting Semi-Supervised Learning with Curriculum Pseudo Labeling (476 citations)
- ABC: Auxiliary Balanced Classifier for Class-imbalanced Semi-supervised Learning (41 citations)
- OpenMatch: Open-Set Semi-supervised Learning with Open-set Consistency Regularization (29 citations)
- RETRIEVE: Coreset Selection for Efficient and Robust Semi-Supervised Learning (17 citations)
- Semi-Supervised Semantic Segmentation via Adaptive Equalization Learning (14 citations)
- TestRank: Bringing Order into Unlabeled Test Instances for Deep Learning Tasks (13 citations)
- Self-Paced Contrastive Learning for Semi-supervised Medical Image Segmentation with Meta-labels (10 citations)
- Combating Noise: Semi-supervised Learning by Region Uncertainty Quantification (9 citations)
- Universal Semi-Supervised Learning (9 citations)
- DP-SSL: Towards Robust Semi-supervised Learning with A Few Labeled Samples (6 citations)
- Adversarial Training Helps Transfer Learning via Better Representations (4 citations)
- Data driven semi-supervised learning (2 citations)
and 2 more in the interactive map.
Knowledge distillation · 14 papers
- Learning curves of generic features maps for realistic datasets with a teacher-student model (55 citations)
- Learning Student-Friendly Teacher Networks for Knowledge Distillation (45 citations)
- Distilling Object Detectors with Feature Richness (36 citations)
- Novel Visual Category Discovery with Dual Ranking Statistics and Mutual Knowledge Distillation (36 citations)
- Adversarial Teacher-Student Representation Learning for Domain Generalization (20 citations)
- Diversity Matters When Learning From Ensembles (16 citations)
- Analyzing the Confidentiality of Undistillable Teachers in Knowledge Distillation (14 citations)
- Comprehensive Knowledge Distillation with Causal Intervention (14 citations)
- Does Knowledge Distillation Really Work? (11 citations)
- Mosaicking to Distill: Knowledge Distillation from Out-of-Domain Data (7 citations)
- Instance-Conditional Knowledge Distillation for Object Detection (6 citations)
- Unsupervised Representation Transfer for Small Networks: I Believe I Can Distill On-the-Fly (6 citations)
and 2 more in the interactive map.
Noisy labels · 11 papers
- Pervasive Label Errors in Test Sets Destabilize Machine Learning Benchmarks (101 citations)
- Understanding and Improving Early Stopping for Learning with Noisy Labels (53 citations)
- FINE Samples for Learning with Noisy Labels (36 citations)
- Instance-dependent Label-noise Learning under a Structural Causal Model (31 citations)
- Open-set Label Noise Can Improve Robustness Against Inherent Label Noise (30 citations)
- Generalized Jensen-Shannon Divergence Loss for Learning with Noisy Labels (22 citations)
- Improved Regularization and Robustness for Fine-tuning in Neural Networks (11 citations)
- Can Less be More? When Increasing-to-Balancing Label Noise Rates Considered Beneficial (3 citations)
- Interactive Label Cleaning with Example-based Explanations (3 citations)
- How does a Neural Network's Architecture Impact its Robustness to Noisy Labels?
- Generalized DataWeighting via Class-Level Gradient Manipulation
Test-time adaptation · 10 papers
- Test-Time Classifier Adjustment Module for Model-Agnostic Domain Generalization (83 citations)
- TTT++: When Does Self-Supervised Test-Time Training Fail or Thrive? (76 citations)
- Adaptive Risk Minimization: Learning to Adapt to Domain Shift (28 citations)
- Test-Time Personalization with a Transformer for Human Pose Estimation (12 citations)
- On Interaction Between Augmentations and Corruptions in Natural Corruption Robustness (10 citations)
- Online Adaptation to Label Distribution Shift (8 citations)
- Bayesian Adaptation for Covariate Shift (6 citations)
- Training for the Future: A Simple Gradient Interpolation Loss to Generalize Along Time (6 citations)
- Cockpit: A Practical Debugging Tool for the Training of Deep Neural Networks (3 citations)
- Characterizing Generalization under Out-Of-Distribution Shifts in Deep Metric Learning (1 citation)
