NeurIPS 2022 accepted papers on Label-efficient and robust learning
345 of the 2,901 papers accepted at NeurIPS 2022 (11.9%) are on Label-efficient and robust learning, compared with 10.4% at NeurIPS 2021. They fall into 9 topics.
See NeurIPS 2022 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 · 161 papers
- Why do tree-based models still outperform deep learning on typical tabular data? (679 citations)
- OpenXAI: Towards a Transparent Evaluation of Model Explanations (81 citations)
- Trajectory-guided Control Prediction for End-to-end Autonomous Driving: A Simple yet Strong Baseline (81 citations)
- Self-Supervised Aggregation of Diverse Experts for Test-Agnostic Long-Tailed Recognition (55 citations)
- The Effects of Regularization and Data Augmentation are Class Dependent (39 citations)
- Concept Embedding Models: Beyond the Accuracy-Explainability Trade-Off (36 citations)
- Diagnosing failures of fairness transfer across distribution shift in real-world medical settings (32 citations)
- Reconstructing Training Data From Trained Neural Networks (28 citations)
- Tenrec: A Large-scale Multipurpose Benchmark Dataset for Recommender Systems (27 citations)
- Improved Fine-Tuning by Better Leveraging Pre-Training Data (24 citations)
- Towards Understanding Grokking: An Effective Theory of Representation Learning (24 citations)
- On Feature Learning in the Presence of Spurious Correlations (22 citations)
and 149 more in the interactive map.
Self-supervised and contrastive · 37 papers
- Masked Autoencoders that Listen (103 citations)
- Toward a realistic model of speech processing in the brain with self-supervised learning (40 citations)
- TransTab: Learning Transferable Tabular Transformers Across Tables (37 citations)
- Self-supervised Heterogeneous Graph Pre-training Based on Structural Clustering (24 citations)
- Contrastive and Non-Contrastive Self-Supervised Learning Recover Global and Local Spectral Embedding Methods (22 citations)
- Self-Supervised Learning of Brain Dynamics from Broad Neuroimaging Data (18 citations)
- Self-Supervised Learning via Maximum Entropy Coding (14 citations)
- Local Spatiotemporal Representation Learning for Longitudinally-consistent Neuroimage Analysis (13 citations)
- MetaMask: Revisiting Dimensional Confounder for Self-Supervised Learning (12 citations)
- Improving Self-Supervised Learning by Characterizing Idealized Representations (8 citations)
- Pre-Train Your Loss: Easy Bayesian Transfer Learning with Informative Priors (8 citations)
- Adapting Self-Supervised Vision Transformers by Probing Attention-Conditioned Masking Consistency (7 citations)
and 25 more in the interactive map.
Domain adaptation and generalization · 36 papers
- Attracting and Dispersing: A Simple Approach for Source-free Domain Adaptation (55 citations)
- Ensemble of Averages: Improving Model Selection and Boosting Performance in Domain Generalization (47 citations)
- Adversarial Style Augmentation for Domain Generalized Urban-Scene Segmentation (36 citations)
- Divide and Contrast: Source-free Domain Adaptation via Adaptive Contrastive Learning (29 citations)
- Unified Optimal Transport Framework for Universal Domain Adaptation (24 citations)
- Domain Generalization by Learning and Removing Domain-specific Features (22 citations)
- Mix and Reason: Reasoning over Semantic Topology with Data Mixing for Domain Generalization (20 citations)
- Sym-NCO: Leveraging Symmetricity for Neural Combinatorial Optimization (20 citations)
- Variational Model Perturbation for Source-Free Domain Adaptation (19 citations)
- Deliberated Domain Bridging for Domain Adaptive Semantic Segmentation (16 citations)
- Few-shot Image Generation via Adaptation-Aware Kernel Modulation (16 citations)
- Meta-DMoE: Adapting to Domain Shift by Meta-Distillation from Mixture-of-Experts (16 citations)
and 24 more in the interactive map.
Zero-/few-shot and meta-learning · 29 papers
- Compositional generalization through abstract representations in human and artificial neural networks (31 citations)
- ReCo: Retrieve and Co-segment for Zero-shot Transfer (29 citations)
- Rethinking Generalization in Few-Shot Classification (29 citations)
- Patching open-vocabulary models by interpolating weights (26 citations)
- Singular Value Fine-tuning: Few-shot Segmentation requires Few-parameters Fine-tuning (26 citations)
- GenerSpeech: Towards Style Transfer for Generalizable Out-Of-Domain Text-to-Speech (21 citations)
- Adaptive Distribution Calibration for Few-Shot Learning with Hierarchical Optimal Transport (12 citations)
- Contrastive Adapters for Foundation Model Group Robustness (12 citations)
- A Closer Look at Prototype Classifier for Few-shot Image Classification (11 citations)
- Few-shot Relational Reasoning via Connection Subgraph Pretraining (9 citations)
- Alleviating the Sample Selection Bias in Few-shot Learning by Removing Projection to the Centroid (8 citations)
- Drawing out of Distribution with Neuro-Symbolic Generative Models (4 citations)
and 17 more in the interactive map.
OOD and anomaly detection · 23 papers
- ADBench: Anomaly Detection Benchmark (252 citations)
- A Unified Model for Multi-class Anomaly Detection (99 citations)
- OpenOOD: Benchmarking Generalized Out-of-Distribution Detection (84 citations)
- Is Out-of-Distribution Detection Learnable? (30 citations)
- SoftPatch: Unsupervised Anomaly Detection with Noisy Data (21 citations)
- Boosting Out-of-distribution Detection with Typical Features (16 citations)
- AnoShift: A Distribution Shift Benchmark for Unsupervised Anomaly Detection (15 citations)
- RankFeat: Rank-1 Feature Removal for Out-of-distribution Detection (14 citations)
- Perturbation Learning Based Anomaly Detection (11 citations)
- Agreement-on-the-line: Predicting the Performance of Neural Networks under Distribution Shift (10 citations)
- Learning Substructure Invariance for Out-of-Distribution Molecular Representations (7 citations)
- UQGAN: A Unified Model for Uncertainty Quantification of Deep Classifiers trained via Conditional GANs (7 citations)
and 11 more in the interactive map.
Knowledge distillation · 22 papers
- Knowledge Distillation from A Stronger Teacher (97 citations)
- Dataset Distillation via Factorization (58 citations)
- PKD: General Distillation Framework for Object Detectors via Pearson Correlation Coefficient (42 citations)
- Bi-directional Weakly Supervised Knowledge Distillation for Whole Slide Image Classification (40 citations)
- Learning Generalizable Models for Vehicle Routing Problems via Knowledge Distillation (25 citations)
- Structural Knowledge Distillation for Object Detection (22 citations)
- Momentum Adversarial Distillation: Handling Large Distribution Shifts in Data-Free Knowledge Distillation (18 citations)
- Improved Feature Distillation via Projector Ensemble (17 citations)
- What Makes a "Good" Data Augmentation in Knowledge Distillation - A Statistical Perspective (16 citations)
- Asymmetric Temperature Scaling Makes Larger Networks Teach Well Again (11 citations)
- Efficient Knowledge Distillation from Model Checkpoints (9 citations)
- Respecting Transfer Gap in Knowledge Distillation (8 citations)
and 10 more in the interactive map.
Semi-supervised learning · 16 papers
- Semi-supervised Semantic Segmentation with Prototype-based Consistency Regularization (43 citations)
- USB: A Unified Semi-supervised Learning Benchmark for Classification (43 citations)
- Debiased Self-Training for Semi-Supervised Learning (39 citations)
- Semi-Supervised Semantic Segmentation via Gentle Teaching Assistant (31 citations)
- Learning from Future: A Novel Self-Training Framework for Semantic Segmentation (21 citations)
- Semi-supervised Vision Transformers at Scale (21 citations)
- Active Learning Through a Covering Lens (12 citations)
- Is one annotation enough? - A data-centric image classification benchmark for noisy and ambiguous label estimation (11 citations)
- Optimizing Data Collection for Machine Learning (11 citations)
- Robust Semi-Supervised Learning when Not All Classes have Labels (9 citations)
- An Embarrassingly Simple Approach to Semi-Supervised Few-Shot Learning (8 citations)
- Few-Shot Non-Parametric Learning with Deep Latent Variable Model (3 citations)
and 4 more in the interactive map.
Test-time adaptation · 11 papers
- Test-Time Training with Masked Autoencoders (36 citations)
- NOTE: Robust Continual Test-time Adaptation Against Temporal Correlation (30 citations)
- Test Time Adaptation via Conjugate Pseudo-labels (24 citations)
- Wild-Time: A Benchmark of in-the-Wild Distribution Shift over Time (22 citations)
- Probable Domain Generalization via Quantile Risk Minimization (16 citations)
- OST: Improving Generalization of DeepFake Detection via One-Shot Test-Time Training (12 citations)
- Robust Calibration with Multi-domain Temperature Scaling (5 citations)
- Task Discovery: Finding the Tasks that Neural Networks Generalize on (4 citations)
- AutoML Two-Sample Test (1 citation)
- MEMO: Test Time Robustness via Adaptation and Augmentation
- Revisiting Realistic Test-Time Training: Sequential Inference and Adaptation by Anchored Clustering
Noisy labels · 10 papers
- Estimating Noise Transition Matrix with Label Correlations for Noisy Multi-Label Learning (6 citations)
- Meta-Query-Net: Resolving Purity-Informativeness Dilemma in Open-set Active Learning (6 citations)
- CS-Shapley: Class-wise Shapley Values for Data Valuation in Classification (5 citations)
- Confidence-based Reliable Learning under Dual Noises (3 citations)
- Learning from Label Proportions by Learning with Label Noise (3 citations)
- Class-Dependent Label-Noise Learning with Cycle-Consistency Regularization (2 citations)
- Noise Attention Learning: Enhancing Noise Robustness by Gradient Scaling (2 citations)
- Lifting Weak Supervision To Structured Prediction (1 citation)
- Robustness to Label Noise Depends on the Shape of the Noise Distribution
- On Image Segmentation With Noisy Labels: Characterization and Volume Properties of the Optimal Solutions to Accuracy and Dice
