ICML 2020 accepted papers on Label-efficient and robust learning
109 of the 1,087 papers accepted at ICML 2020 (10.0%) are on Label-efficient and robust learning, compared with 8.4% at ICML 2019. They fall into 9 topics.
See ICML 2020 in the interactive map
Submitting to ICML? 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 · 49 papers
- Understanding Contrastive Representation Learning through Alignment and Uniformity on the Hypersphere (168 citations)
- Concept Bottleneck Models (149 citations)
- Interpretations are Useful: Penalizing Explanations to Align Neural Networks with Prior Knowledge (98 citations)
- Learning De-biased Representations with Biased Representations (95 citations)
- Weakly-Supervised Disentanglement Without Compromises (90 citations)
- Aligned Cross Entropy for Non-Autoregressive Machine Translation (78 citations)
- Revisiting Training Strategies and Generalization Performance in Deep Metric Learning (78 citations)
- LEEP: A New Measure to Evaluate Transferability of Learned Representations (69 citations)
- From ImageNet to Image Classification: Contextualizing Progress on Benchmarks (64 citations)
- Confidence-Aware Learning for Deep Neural Networks (45 citations)
- Unsupervised Speech Decomposition via Triple Information Bottleneck (42 citations)
- Description Based Text Classification with Reinforcement Learning (40 citations)
and 37 more in the interactive map.
Domain adaptation and generalization · 17 papers
- Do We Really Need to Access the Source Data? Source Hypothesis Transfer for Unsupervised Domain Adaptation (140 citations)
- Efficient Domain Generalization via Common-Specific Low-Rank Decomposition (68 citations)
- Implicit Class-Conditioned Domain Alignment for Unsupervised Domain Adaptation (50 citations)
- Data Valuation using Reinforcement Learning (45 citations)
- Progressive Graph Learning for Open-Set Domain Adaptation (43 citations)
- Few-shot Domain Adaptation by Causal Mechanism Transfer (35 citations)
- Domain Aggregation Networks for Multi-Source Domain Adaptation (27 citations)
- Angular Visual Hardness (19 citations)
- Estimating Generalization under Distribution Shifts via Domain-Invariant Representations (19 citations)
- Understanding Self-Training for Gradual Domain Adaptation (19 citations)
- Label-Noise Robust Domain Adaptation (17 citations)
- LTF: A Label Transformation Framework for Correcting Label Shift (16 citations)
and 5 more in the interactive map.
Self-supervised and contrastive · 12 papers
- A Simple Framework for Contrastive Learning of Visual Representations (7,181 citations)
- PEGASUS: Pre-training with Extracted Gap-sentences for Abstractive Summarization (975 citations)
- Data-Efficient Image Recognition with Contrastive Predictive Coding (906 citations)
- Generative Pretraining From Pixels (549 citations)
- Unsupervised Discovery of Interpretable Directions in the GAN Latent Space (184 citations)
- Graph-based, Self-Supervised Program Repair from Diagnostic Feedback (63 citations)
- Self-supervised Label Augmentation via Input Transformations (54 citations)
- InfoGAN-CR and ModelCentrality: Self-supervised Model Training and Selection for Disentangling GANs (43 citations)
- Automatic Shortcut Removal for Self-Supervised Representation Learning (40 citations)
- Distribution Augmentation for Generative Modeling (26 citations)
- A Sequential Self Teaching Approach for Improving Generalization in Sound Event Recognition (20 citations)
- Learning Compound Tasks without Task-specific Knowledge via Imitation and Self-supervised Learning (8 citations)
Noisy labels · 10 papers
- Normalized Loss Functions for Deep Learning with Noisy Labels (122 citations)
- Beyond Synthetic Noise: Deep Learning on Controlled Noisy Labels (116 citations)
- Peer Loss Functions: Learning from Noisy Labels without Knowing Noise Rates (97 citations)
- Error-Bounded Correction of Noisy Labels (36 citations)
- Training Binary Neural Networks through Learning with Noisy Supervision (30 citations)
- Does label smoothing mitigate label noise? (16 citations)
- Improving generalization by controlling label-noise information in neural network weights (6 citations)
- Feature Noise Induces Loss Discrepancy Across Groups (4 citations)
- Deep k-NN for Noisy Labels (2 citations)
- Learning with Bounded Instance and Label-dependent Label Noise
Zero-/few-shot and meta-learning · 7 papers
- Laplacian Regularized Few-Shot Learning (55 citations)
- TaskNorm: Rethinking Batch Normalization for Meta-Learning (30 citations)
- MetaFun: Meta-Learning with Iterative Functional Updates (29 citations)
- Unraveling Meta-Learning: Understanding Feature Representations for Few-Shot Tasks (16 citations)
- Hallucinative Topological Memory for Zero-Shot Visual Planning (11 citations)
- XtarNet: Learning to Extract Task-Adaptive Representation for Incremental Few-Shot Learning (7 citations)
- Meta Variance Transfer: Learning to Augment from the Others
Semi-supervised learning · 5 papers
- Safe Deep Semi-Supervised Learning for Unseen-Class Unlabeled Data (79 citations)
- Semi-Supervised StyleGAN for Disentanglement Learning (48 citations)
- Poisson Learning: Graph Based Semi-Supervised Learning At Very Low Label Rates (28 citations)
- Time-Consistent Self-Supervision for Semi-Supervised Learning (16 citations)
- Negative Sampling in Semi-Supervised learning (8 citations)
OOD and anomaly detection · 4 papers
- Detecting Out-of-Distribution Examples with Gram Matrices (67 citations)
- Can Autonomous Vehicles Identify, Recover From, and Adapt to Distribution Shifts? (63 citations)
- Interpretable, Multidimensional, Multimodal Anomaly Detection with Negative Sampling for Detection of Device Failure (14 citations)
- NADS: Neural Architecture Distribution Search for Uncertainty Awareness (10 citations)
