ICLR 2022 accepted papers on Label-efficient and robust learning
152 of the 1,095 papers accepted at ICLR 2022 (13.9%) are on Label-efficient and robust learning, compared with 16.2% at ICLR 2021. They fall into 9 topics.
See ICLR 2022 in the interactive map
Submitting to ICLR? 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 · 69 papers
- Uncertainty Modeling for Out-of-Distribution Generalization (71 citations)
- PiCO: Contrastive Label Disambiguation for Partial Label Learning (55 citations)
- Open-Set Recognition: A Good Closed-Set Classifier is All You Need (40 citations)
- Adversarial Retriever-Ranker for Dense Text Retrieval (36 citations)
- SphereFace2: Binary Classification is All You Need for Deep Face Recognition (25 citations)
- Enhancing Cross-lingual Transfer by Manifold Mixup (23 citations)
- Post hoc Explanations may be Ineffective for Detecting Unknown Spurious Correlation (21 citations)
- CADDA: Class-wise Automatic Differentiable Data Augmentation for EEG Signals (17 citations)
- Discovering Latent Concepts Learned in BERT (17 citations)
- DISCOVERING AND EXPLAINING THE REPRESENTATION BOTTLENECK OF DNNS (16 citations)
- MetaShift: A Dataset of Datasets for Evaluating Contextual Distribution Shifts and Training Conflicts (13 citations)
- Spread Spurious Attribute: Improving Worst-group Accuracy with Spurious Attribute Estimation (11 citations)
and 57 more in the interactive map.
Self-supervised and contrastive · 23 papers
- VICReg: Variance-Invariance-Covariance Regularization for Self-Supervised Learning (335 citations)
- Understanding Dimensional Collapse in Contrastive Self-supervised Learning (91 citations)
- Efficient Self-supervised Vision Transformers for Representation Learning (82 citations)
- Self-supervised Learning is More Robust to Dataset Imbalance (64 citations)
- Zero-Shot Self-Supervised Learning for MRI Reconstruction (31 citations)
- Incremental False Negative Detection for Contrastive Learning (25 citations)
- How Does SimSiam Avoid Collapse Without Negative Samples? A Unified Understanding with Self-supervised Contrastive Learning (23 citations)
- Chaos is a Ladder: A New Theoretical Understanding of Contrastive Learning via Augmentation Overlap (17 citations)
- NODE-GAM: Neural Generalized Additive Model for Interpretable Deep Learning (14 citations)
- Optimal Representations for Covariate Shift (11 citations)
- How Well Does Self-Supervised Pre-Training Perform with Streaming Data? (10 citations)
- Scarf: Self-Supervised Contrastive Learning using Random Feature Corruption (10 citations)
and 11 more in the interactive map.
Domain adaptation and generalization · 19 papers
- CDTrans: Cross-domain Transformer for Unsupervised Domain Adaptation (111 citations)
- Leveraging unlabeled data to predict out-of-distribution performance (19 citations)
- Gradient Matching for Domain Generalization (13 citations)
- Extending the WILDS Benchmark for Unsupervised Adaptation (12 citations)
- Discrepancy-Based Active Learning for Domain Adaptation (11 citations)
- Mapping conditional distributions for domain adaptation under generalized target shift (7 citations)
- Adversarial Support Alignment (3 citations)
- Controlling Directions Orthogonal to a Classifier (3 citations)
- Source-Free Adaptation to Measurement Shift via Bottom-Up Feature Restoration (3 citations)
- Provable Adaptation across Multiway Domains via Representation Learning (2 citations)
- Auto-Transfer: Learning to Route Transferable Representations (1 citation)
- From Intervention to Domain Transportation: A Novel Perspective to Optimize Recommendation (1 citation)
and 7 more in the interactive map.
Zero-/few-shot and meta-learning · 13 papers
- Multitask Prompted Training Enables Zero-Shot Task Generalization (556 citations)
- How to Train Your MAML to Excel in Few-Shot Classification (16 citations)
- Meta Discovery: Learning to Discover Novel Classes given Very Limited Data (16 citations)
- Meta-Learning with Fewer Tasks through Task Interpolation (11 citations)
- On the Role of Neural Collapse in Transfer Learning (11 citations)
- Towards Better Understanding and Better Generalization of Low-shot Classification in Histology Images with Contrastive Learning (11 citations)
- MAML is a Noisy Contrastive Learner in Classification (5 citations)
- Learning Prototype-oriented Set Representations for Meta-Learning (4 citations)
- On the Importance of Firth Bias Reduction in Few-Shot Classification (4 citations)
- Hierarchical Variational Memory for Few-shot Learning Across Domains (3 citations)
- Task Affinity with Maximum Bipartite Matching in Few-Shot Learning (1 citation)
- Closed-form Sample Probing for Learning Generative Models in Zero-shot Learning
and 1 more in the interactive map.
Semi-supervised learning · 9 papers
- Open-World Semi-Supervised Learning (45 citations)
- Backdoor Defense via Decoupling the Training Process (41 citations)
- AdaMatch: A Unified Approach to Semi-Supervised Learning and Domain Adaptation (22 citations)
- Pseudo-Labeled Auto-Curriculum Learning for Semi-Supervised Keypoint Localization (8 citations)
- The Rich Get Richer: Disparate Impact of Semi-Supervised Learning (8 citations)
- Surreal-GAN:Semi-Supervised Representation Learning via GAN for uncovering heterogeneous disease-related imaging patterns (6 citations)
- Learning to Annotate Part Segmentation with Gradient Matching (3 citations)
- On Non-Random Missing Labels in Semi-Supervised Learning (3 citations)
- How unlabeled data improve generalization in self-training? A one-hidden-layer theoretical analysis (1 citation)
OOD and anomaly detection · 6 papers
- Igeood: An Information Geometry Approach to Out-of-Distribution Detection (11 citations)
- PI3NN: Out-of-distribution-aware Prediction Intervals from Three Neural Networks (8 citations)
- A Statistical Framework for Efficient Out of Distribution Detection in Deep Neural Networks (1 citation)
- Meta Learning Low Rank Covariance Factors for Energy Based Deterministic Uncertainty
- Anomaly Detection for Tabular Data with Internal Contrastive Learning
- Asymmetry Learning for Counterfactually-invariant Classification in OOD Tasks
Noisy labels · 5 papers
- Learning with Noisy Labels Revisited: A Study Using Real-World Human Annotations (60 citations)
- Sample Selection with Uncertainty of Losses for Learning with Noisy Labels (47 citations)
- Noisy Feature Mixup
- An Information Fusion Approach to Learning with Instance-Dependent Label Noise
- Resolving Training Biases via Influence-based Data Relabeling
Knowledge distillation · 4 papers
Test-time adaptation · 4 papers
- Visual Representation Learning Does Not Generalize Strongly Within the Same Domain (6 citations)
- Learning to Generalize across Domains on Single Test Samples (5 citations)
- Distributionally Robust Models with Parametric Likelihood Ratios (2 citations)
- Tracking the risk of a deployed model and detecting harmful distribution shifts (2 citations)
