Learning to Reweight Examples for Robust Deep Learning

Abstract

Deep neural networks have been shown to be very powerful modeling tools for many supervised learning tasks involving complex input patterns. However, they can also easily overfit to training set biases and label noises. In addition to various regularizers, example reweighting algorithms are popular solutions to these problems, but they require careful tuning of additional hyperparameters, such as example mining schedules and regularization hyperparameters. In contrast to past reweighting methods, which typically consist of functions of the cost value of each example, in this work we propose a novel meta-learning algorithm that learns to assign weights to training examples based on their gradient directions. To determine the example weights, our method performs a meta gradient descent step on the current mini-batch example weights (which are initialized from zero) to minimize the loss on a clean unbiased validation set. Our proposed method can be easily implemented on any type of deep network, does not require any additional hyperparameter tuning, and achieves impressive performance on class imbalance and corrupted label problems where only a small amount of clean validation data is available.

13 pages; Published at ICML 2018; Code released at: https://github.com/uber-research/learning-to-reweight-examples

References in corpus (2)

• Training Deep Neural Networks on Noisy Labels with Bootstrapping
• Stochastic Hyperparameter Optimization through Hypernetworks

Cited by in corpus (193)

• Generalized Cross Entropy Loss for Training Deep Neural Networks with Noisy Labels
• Deep Multi-modal Object Detection and Semantic Segmentation for Autonomous Driving: Datasets, Methods, and Challenges
• Regularized Deep Networks in Intelligent Transportation Systems: A Taxonomy and a Case Study
• Co-teaching: Robust Training of Deep Neural Networks with Extremely Noisy Labels
• Meta-Weight-Net: Learning an Explicit Mapping For Sample Weighting
• Image Classification with Deep Learning in the Presence of Noisy Labels: A Survey
• Balanced Meta-Softmax for Long-Tailed Visual Recognition
• Early-Learning Regularization Prevents Memorization of Noisy Labels
• Are Anchor Points Really Indispensable in Label-Noise Learning?
• Masking: A New Perspective of Noisy Supervision
• Gradient Descent with Early Stopping is Provably Robust to Label Noise for Overparameterized Neural Networks
• Robust Federated Learning with Noisy Labels
• Using Trusted Data to Train Deep Networks on Labels Corrupted by Severe Noise
• LongReMix: Robust Learning with High Confidence Samples in a Noisy Label Environment
• Dual T: Reducing Estimation Error for Transition Matrix in Label-noise Learning
• Toward Open-World Electroencephalogram Decoding Via Deep Learning: A Comprehensive Survey
• Long-tailed Visual Recognition via Gaussian Clouded Logit Adjustment
• Towards Intersectionality in Machine Learning: Including More Identities, Handling Underrepresentation, and Performing Evaluation
• Distribution Aligning Refinery of Pseudo-label for Imbalanced Semi-supervised Learning
• Self-Adaptive Training: beyond Empirical Risk Minimization
• Coresets via Bilevel Optimization for Continual Learning and Streaming
• Learning to Generalize from Sparse and Underspecified Rewards
• Meta Balanced Network for Fair Face Recognition
• Combating noisy labels by agreement: A joint training method with co-regularization
• Self-supervised Learning is More Robust to Dataset Imbalance
• Co-Correcting: Noise-tolerant Medical Image Classification via mutual Label Correction
• SELF: Learning to Filter Noisy Labels with Self-Ensembling
• Identifying and Compensating for Feature Deviation in Imbalanced Deep Learning
• Long-Tailed Classification by Keeping the Good and Removing the Bad Momentum Causal Effect
• On the Minimal Supervision for Training Any Binary Classifier from Only Unlabeled Data
• Feature Fusion from Head to Tail for Long-Tailed Visual Recognition
• Denoising Multi-Source Weak Supervision for Neural Text Classification
• Identifying Mislabeled Data using the Area Under the Margin Ranking
• IMAE for Noise-Robust Learning: Mean Absolute Error Does Not Treat Examples Equally and Gradient Magnitude's Variance Matters
• An Entropic Optimal Transport Loss for Learning Deep Neural Networks under Label Noise in Remote Sensing Images
• Label Noise Types and Their Effects on Deep Learning
• Balancing Unobserved Confounding with a Few Unbiased Ratings in Debiased Recommendations
• Learning from Pixel-Level Label Noise: A New Perspective for Semi-Supervised Semantic Segmentation
• FINE Samples for Learning with Noisy Labels
• Distribution-Balanced Loss for Multi-Label Classification in Long-Tailed Datasets
• Adaptive Self-training for Few-shot Neural Sequence Labeling
• CReST: A Class-Rebalancing Self-Training Framework for Imbalanced Semi-Supervised Learning
• L_DMI: An Information-theoretic Noise-robust Loss Function
• Boosting Facial Expression Recognition by A Semi-Supervised Progressive Teacher
• Open-set Label Noise Can Improve Robustness Against Inherent Label Noise
• Deep Imbalanced Learning for Face Recognition and Attribute Prediction
• Rethinking Class-Balanced Methods for Long-Tailed Visual Recognition from a Domain Adaptation Perspective
• The DNNLikelihood: enhancing likelihood distribution with Deep Learning
• MESA: Boost Ensemble Imbalanced Learning with MEta-SAmpler
• Provably End-to-end Label-Noise Learning without Anchor Points
• TGDM: Target Guided Dynamic Mixup for Cross-Domain Few-Shot Learning
• Uncertainty Based Detection and Relabeling of Noisy Image Labels
• Penalty Method for Inversion-Free Deep Bilevel Optimization
• Not All Unlabeled Data are Equal: Learning to Weight Data in Semi-supervised Learning
• Hyperbolic Geometric Graph Representation Learning for Hierarchy-imbalance Node Classification
• Heteroskedastic and Imbalanced Deep Learning with Adaptive Regularization
• Distribution Density, Tails, and Outliers in Machine Learning: Metrics and Applications
• MetaLabelNet: Learning to Generate Soft-Labels from Noisy-Labels
• M2m: Imbalanced Classification via Major-to-minor Translation
• Learning Soft Labels via Meta Learning
• How does Early Stopping Help Generalization against Label Noise?
• Feature-Balanced Loss for Long-Tailed Visual Recognition
• MetaSAug: Meta Semantic Augmentation for Long-Tailed Visual Recognition
• Robust Learning Under Label Noise With Iterative Noise-Filtering
• How to Train Your MAML to Excel in Few-Shot Classification
• Deep Self-Learning From Noisy Labels
• SimAug: Learning Robust Representations from Simulation for Trajectory Prediction
• Simple and Effective Regularization Methods for Training on Noisily Labeled Data with Generalization Guarantee
• SWIPENET: Object detection in noisy underwater images
• Exploring Memorization in Adversarial Training
• ProSelfLC: Progressive Self Label Correction for Training Robust Deep Neural Networks
• Learning Bounds for Risk-sensitive Learning
• FR-Train: A Mutual Information-Based Approach to Fair and Robust Training
• Co-Learning Meets Stitch-Up for Noisy Multi-label Visual Recognition
• Rethinking Importance Weighting for Deep Learning under Distribution Shift
• Narrowing the Gap: Improved Detector Training with Noisy Location Annotations
• Balancing Training for Multilingual Neural Machine Translation
• Underwater object detection using Invert Multi-Class Adaboost with deep learning
• Improving Medical Image Classification with Label Noise Using Dual-uncertainty Estimation
• Detecting, Localising and Classifying Polyps from Colonoscopy Videos using Deep Learning
• On Tilted Losses in Machine Learning: Theory and Applications
• Consensual Collaborative Training And Knowledge Distillation Based Facial Expression Recognition Under Noisy Annotations
• Searching to Exploit Memorization Effect in Learning from Corrupted Labels
• Affect Expression Behaviour Analysis in the Wild using Consensual Collaborative Training
• Adversarial Robustness under Long-Tailed Distribution
• Imbalanced Continual Learning with Partitioning Reservoir Sampling
• A Survey on Curriculum Learning
• Derivative Manipulation for General Example Weighting
• Universal Stagewise Learning for Non-Convex Problems with Convergence on Averaged Solutions
• Learning to Impute: A General Framework for Semi-supervised Learning
• Exploiting All Samples in Low-Resource Sentence Classification: Early Stopping and Initialization Parameters
• DoubleEnsemble: A New Ensemble Method Based on Sample Reweighting and Feature Selection for Financial Data Analysis
• Teacher Supervises Students How to Learn From Partially Labeled Images for Facial Landmark Detection
• Task Agnostic Robust Learning on Corrupt Outputs by Correlation-Guided Mixture Density Networks
• Faster Meta Update Strategy for Noise-Robust Deep Learning
• Combining Self-Supervised and Supervised Learning with Noisy Labels
• Label-Noise Robust Generative Adversarial Networks
• Improving Generalization by Controlling Label-Noise Information in Neural Network Weights
• MetaMixUp: Learning Adaptive Interpolation Policy of MixUp with Meta-Learning
• Imbalanced Image Classification with Complement Cross Entropy
• Self-Adaptive Training: Bridging Supervised and Self-Supervised Learning
• Countering Noisy Labels By Learning From Auxiliary Clean Labels
• Adaptive Sample Selection for Robust Learning under Label Noise
• Robust and On-the-fly Dataset Denoising for Image Classification
• Weakly Supervised Learning with Side Information for Noisy Labeled Images
• Few-Shot Anomaly Detection for Polyp Frames from Colonoscopy
• Improving Robustness of Learning-based Autonomous Steering Using Adversarial Images
• The Resistance to Label Noise in K-NN and DNN Depends on its Concentration
• A Novel Self-Supervised Re-labeling Approach for Training with Noisy Labels
• Teaching with Commentaries
• Select-ProtoNet: Learning to Select for Few-Shot Disease Subtype Prediction
• Discriminative-Generative Dual Memory Video Anomaly Detection
• Tilted Empirical Risk Minimization
• Optimizing Data Usage via Differentiable Rewards
• Deep Learning Classification With Noisy Labels
• Cavity Filling: Pseudo-Feature Generation for Multi-Class Imbalanced Data Problems in Deep Learning
• Learning Image Labels On-the-fly for Training Robust Classification Models
• Deep k-NN for Noisy Labels
• Less Is Better: Unweighted Data Subsampling via Influence Function
• How Out-of-Distribution Data Hurts Semi-Supervised Learning
• Learning to Combat Noisy Labels via Classification Margins
• Joint Negative and Positive Learning for Noisy Labels
• Robust Deep Graph Based Learning for Binary Classification
• NLNL: Negative Learning for Noisy Labels
• Using Latent Codes for Class Imbalance Problem in Unsupervised Domain Adaptation
• Safeguarded Dynamic Label Regression for Generalized Noisy Supervision
• Reducing and Exploiting Data Augmentation Noise through Meta Reweighting Contrastive Learning for Text Classification
• Online Hyperparameter Meta-Learning with Hypergradient Distillation
• Data Manipulation: Towards Effective Instance Learning for Neural Dialogue Generation via Learning to Augment and Reweight
• No Regret Sample Selection with Noisy Labels
• Learning with Noisy Labels by Efficient Transition Matrix Estimation to Combat Label Miscorrection
• A Simple yet Effective Baseline for Robust Deep Learning with Noisy Labels
• Improving Medical Annotation Quality to Decrease Labeling Burden Using Stratified Noisy Cross-Validation
• Unified Gradient Reweighting for Model Biasing with Applications to Source Separation
• From Recognition to Prediction: Analysis of Human Action and Trajectory Prediction in Video
• Optimizing Black-box Metrics with Iterative Example Weighting
• MetaPerturb: Transferable Regularizer for Heterogeneous Tasks and Architectures
• Direct Differentiable Augmentation Search
• Negative-ResNet: Noisy Ambulatory Electrocardiogram Signal Classification Scheme
• Distilling effective supervision for robust medical image segmentation with noisy labels
• Adversarial-Based Knowledge Distillation for Multi-Model Ensemble and Noisy Data Refinement
• Minimizing Close-k Aggregate Loss Improves Classification
• Class-Wise Difficulty-Balanced Loss for Solving Class-Imbalance
• Meta Label Correction for Noisy Label Learning
• Self-Paced Deep Regression Forests with Consideration on Underrepresented Examples
• Meta Corrupted Pixels Mining for Medical Image Segmentation
• Few-Shot Text Ranking with Meta Adapted Synthetic Weak Supervision
• Leveraging Local Variation in Data: Sampling and Weighting Schemes for Supervised Deep Learning
• Energy Aligning for Biased Models
• Medical Image Segmentation with Limited Supervision: A Review of Deep Network Models
• Mobility-aware Content Preference Learning in Decentralized Caching Networks
• Learning to Auto Weight: Entirely Data-driven and Highly Efficient Weighting Framework
• ARM: A Confidence-Based Adversarial Reweighting Module for Coarse Semantic Segmentation
• Self-Supervised Person Detection in 2D Range Data using a Calibrated Camera
• Constrained Instance and Class Reweighting for Robust Learning under Label Noise
• Improving Training on Noisy Stuctured Labels
• Graph convolutional networks for learning with few clean and many noisy labels
• Multiplicative Reweighting for Robust Neural Network Optimization
• Sample Prior Guided Robust Model Learning to Suppress Noisy Labels
• MULTIMODAL ANALYSIS: Informed content estimation and audio source separation
• Procrustean Training for Imbalanced Deep Learning
• Learning to Transfer Learn: Reinforcement Learning-Based Selection for Adaptive Transfer Learning
• Learning from Web Data with Self-Organizing Memory Module
• Ada-Segment: Automated Multi-loss Adaptation for Panoptic Segmentation
• Exploiting Class Similarity for Machine Learning with Confidence Labels and Projective Loss Functions
• Self-Training with Weak Supervision
• Learning to Reweight with Deep Interactions
• Task-Driven Data Verification via Gradient Descent
• Semi-Supervised Learning with Meta-Gradient
• Spending Money Wisely: Online Electronic Coupon Allocation based on Real-Time User Intent Detection
• Two-Stream Compare and Contrast Network for Vertebral Compression Fracture Diagnosis
• Gradient-based Hyperparameter Optimization Over Long Horizons
• Improving Generalization of Deep Fault Detection Models in the Presence of Mislabeled Data
• MSD: Saliency-aware Knowledge Distillation for Multimodal Understanding
• Which Samples Should be Learned First: Easy or Hard?
• DIVA: Dataset Derivative of a Learning Task
• Learning from Mistakes based on Class Weighting with Application to Neural Architecture Search
• One-shot Weakly-Supervised Segmentation in Medical Images
• Learning advisor networks for noisy image classification
• Simple and Effective Input Reformulations for Translation
• Compensation Learning
• Biquality Learning: a Framework to Design Algorithms Dealing with Closed-Set Distribution Shifts
• Boosting Mapping Functionality of Neural Networks via Latent Feature Generation based on Reversible Learning
• Meta Adaptation using Importance Weighted Demonstrations
• Sum of Ranked Range Loss for Supervised Learning
• Noise Robust Generative Adversarial Networks
• Temporal-aware Language Representation Learning From Crowdsourced Labels
• Meta Auxiliary Learning for Facial Action Unit Detection
• IPOF: An Extremely and Excitingly Simple Outlier Detection Booster via Infinite Propagation
• Learning with Noisy Labels for Sentence-level Sentiment Classification
• Calibrating Class Activation Maps for Long-Tailed Visual Recognition
• Mitigating Class Boundary Label Uncertainty to Reduce Both Model Bias and Variance
• Robust Temporal Ensembling for Learning with Noisy Labels