ICML-20 待读的 Paper

2020.9.12
花了一上午的时间，过了一遍 ICML-2020 Accepted Paper List, 挑出了自己想读的 Paper。
主要关注于自己的一些研究点。

Noisy Labels

• Beyond Synthetic Noise: Deep Learning on Controlled Noisy Labels
• SIGUA: Forgetting May Make Learning with Noisy Labels More Robust
• Error-Bounded Correction of Noisy Labels
• Does label smoothing mitigate label noise?
• Deep k-NN for Noisy Labels
• Improving generalization by controlling label-noise information in neural network weights
• Normalized Loss Functions for Deep Learning with Noisy Labels
• Variational Label Enhancement
• Peer Loss Functions: Learning from Noisy Labels without Knowing Noise Rates
• Searching to Exploit Memorization Effect in Learning with Noisy Labels
• Label-Noise Robust Domain Adaptation
• Training Binary Neural Networks through Learning with Noisy Supervision
• Learning with Bounded Instance- and Label-dependent Label Noise
• Progressive Identification of True Labels for Partial-Label Learning
• Learning with Multiple Complementary Labels
• Unbiased Risk Estimators Can Mislead: A Case Study of Learning with Complementary Labels

Semi-Supervised Learning

• Semi-Supervised Learning with Normalizing Flows
• Negative Sampling in Semi-Supervised learning
• Poisson Learning: Graph Based Semi-Supervised Learning At Very Low Label Rates
• Time-Consistent Self-Supervision for Semi-Supervised Learning
• Safe Deep Semi-Supervised Learning for Unseen-Class Unlabeled Data
• Deep Streaming Label Learning

Domain Adaptation

• Continuously Indexed Domain Adaptation
• RIFLE: Backpropagation in Depth for Deep Transfer Learning through Re-Initializing the Fully-connected LayEr
• Implicit Class-Conditioned Domain Alignment for Unsupervised Domain Adaptation
• Understanding Self-Training for Gradual Domain Adaptation
• Do We Really Need to Access the Source Data? Source Hypothesis Transfer for Unsupervised Domain Adaptation
• LTF: A Label Transformation Framework for Correcting Label Shift
• Label-Noise Robust Domain Adaptation

Data Bias, Weighting

• Adversarial Filters of Dataset Biases
• Optimizing Data Usage via Differentiable Rewards
• Data preprocessing to mitigate bias: A maximum entropy based approach
• DeBayes: a Bayesian Method for Debiasing Network Embeddings
• A Distributional Framework For Data Valuation

Class-Imbalance

• Class-Weighted Classification: Trade-offs and Robust Approaches
• Online Continual Learning from Imbalanced Data
• Logistic Regression for Massive Data with Rare Events

MixUp, Interpolation, Extrapolation, etc.

• Puzzle Mix: Exploiting Saliency and Local Statistics for Optimal Mixup
• Learning Representations that Support Extrapolation
• Extrapolation for Large-batch Training in Deep Learning
• Training Neural Networks for and by Interpolation
• Sequence Generation with Mixed Representations

PU Learning

• Self-PU: Self Boosted and Calibrated Positive-Unlabeled Training

Active Learning

• Adaptive Region-Based Active Learning

GCN or Recommendation System

• Continuous Graph Neural Networks
• Simple and Deep Graph Convolutional Networks
• Graph Convolutional Network for Recommendation with Low-pass Collaborative Filters
• Generalization and Representational Limits of Graph Neural Networks
• Graph-based Nearest Neighbor Search: From Practice to Theory
• Ordinal Non-negative Matrix Factorization for Recommendation
• Improved Communication Cost in Distributed PageRank Computation – A Theoretical Study
• Optimization and Analysis of the pAp@k Metric for Recommender Systems
• Scalable and Efficient Comparison-based Search without Features
• Learning to Rank Learning Curves
• When Does Self-Supervision Help Graph Convolutional Networks?

Neural ODE

• Towards Adaptive Residual Network Training: A Neural-ODE Perspective
• Combining Differentiable PDE Solvers and Graph Neural Networks for Fluid Flow Prediction
• Approximation Capabilities of Neural ODEs and Invertible Residual Networks

Calibration, Confidence, Out-of-distribution

• Confidence-Aware Learning for Deep Neural Networks
• Being Bayesian, Even Just a Bit, Fixes Overconfidence in ReLU Networks
• SDE-Net: Equipping Deep Neural Networks with Uncertainty Estimates
• Detecting Out-of-Distribution Examples with Gram Matrices
• Interpretable, Multidimensional, Multimodal Anomaly Detection with Negative Sampling for Detection of Device Failure
• Uncertainty Estimation Using a Single Deep Deterministic Neural Network
• How Good is the Bayes Posterior in Deep Neural Networks Really?

Federated Learning, Fairness

• Fair k-Centers via Maximum Matching
• Federated Learning with Only Positive Labels

Interesting Problems, Settings

• Why Are Learned Indexes So Effective?
• Learning with Feature and Distribution Evolvable Streams
• Cost-effectively Identifying Causal Effects When Only Response Variable is Observable
• Rigging the Lottery: Making All Tickets Win
• Do We Need Zero Training Loss After Achieving Zero Training Error?
• Small Data, Big Decisions: Model Selection in the Small-Data Regime
• Why bigger is not always better: on finite and infinite neural networks
• On Learning Sets of Symmetric Elements
• Collaborative Machine Learning with Incentive-Aware Model Rewards
• Generalisation error in learning with random features and the hidden manifold model
• Sample Amplification: Increasing Dataset Size even when Learning is Impossible
• When are Non-Parametric Methods Robust?
• Performative Prediction
• Supervised learning: no loss no cry
• Teaching with Limited Information on the Learner's Behaviour
• Learning De-biased Representations with Biased Representations
• Do RNN and LSTM have Long Memory?
• It's Not What Machines Can Learn, It's What We Cannot Teach
• Enhancing Simple Models by Exploiting What They Already Know

Interesting Theory

• On the Generalization Benefit of Noise in Stochastic Gradient Descent
• Let's Agree to Agree: Neural Networks Share Classification Order on Real Datasets
• Optimizer Benchmarking Needs to Account for Hyperparameter Tuning
• On the Noisy Gradient Descent that Generalizes as SGD
• Rethinking Bias-Variance Trade-off for Generalization of Neural Networks
• Understanding and Mitigating the Tradeoff between Robustness and Accuracy
• The Implicit and Explicit Regularization Effects of Dropout
• Optimal Continual Learning has Perfect Memory and is NP-hard
• Curvature-corrected learning dynamics in deep neural networks
• Explainable k-Means and k-Medians Clustering
• Near-Tight Margin-Based Generalization Bounds for Support Vector Machines
• Uniform Convergence of Rank-weighted Learning
• Decision Trees for Decision-Making under the Predict-then-Optimize Framework

Interesting Algorithm

• SoftSort: A Continuous Relaxation for the argsort Operator
• Boosting Deep Neural Network Efficiency with Dual-Module Inference
• Circuit-Based Intrinsic Methods to Detect Overfitting
• Learning Similarity Metrics for Numerical Simulations
• Deep Divergence Learning
• Consistent Estimators for Learning to Defer to an Expert
• Smaller, more accurate regression forests using tree alternating optimization
• Learning To Stop While Learning To Predict
• DROCC: Deep Robust One-Class Classification

Point Cloud, 3 dimension

• Hypernetwork approach to generating point clouds