An Empirical Study of Example Forgetting during Deep Neural Network Learning

Abstract

Inspired by the phenomenon of catastrophic forgetting, we investigate the learning dynamics of neural networks as they train on single classification tasks. Our goal is to understand whether a related phenomenon occurs when data does not undergo a clear distributional shift. We define a `forgetting event' to have occurred when an individual training example transitions from being classified correctly to incorrectly over the course of learning. Across several benchmark data sets, we find that: (i) certain examples are forgotten with high frequency, and some not at all; (ii) a data set's (un)forgettable examples generalize across neural architectures; and (iii) based on forgetting dynamics, a significant fraction of examples can be omitted from the training data set while still maintaining state-of-the-art generalization performance.

arXiv commentICLR 2019

Cited by in corpus (42)

• Continual Learning for Recurrent Neural Networks: an Empirical Evaluation
• Active label cleaning for improved dataset quality under resource constraints
• What Neural Networks Memorize and Why: Discovering the Long Tail via Influence Estimation
• Selection via Proxy: Efficient Data Selection for Deep Learning
• Dataset Condensation with Differentiable Siamese Augmentation
• Data Valuation using Reinforcement Learning
• MEST: Accurate and Fast Memory-Economic Sparse Training Framework on the Edge
• Orthogonal Gradient Descent for Continual Learning
• Learning from Noisy Labels for Entity-Centric Information Extraction
• Active Domain Randomization
• Federated Few-Shot Learning for Mobile NLP
• Improved Schemes for Episodic Memory-based Lifelong Learning
• The Evolution of Out-of-Distribution Robustness Throughout Fine-Tuning
• Deep Learning Through the Lens of Example Difficulty
• Meta-Learning with Sparse Experience Replay for Lifelong Language Learning
• GRAD-MATCH: Gradient Matching based Data Subset Selection for Efficient Deep Model Training
• Robust Training with Ensemble Consensus
• Topology-based Representative Datasets to Reduce Neural Network Training Resources
• When Do Curricula Work?
• Characterizing Structural Regularities of Labeled Data in Overparameterized Models
• Empirical Study of Easy and Hard Examples in CNN Training
• On the Importance of Difficulty Calibration in Membership Inference Attacks
• SSSE: Efficiently Erasing Samples from Trained Machine Learning Models
• Automatic Construction of Multi-layer Perceptron Network from Streaming Examples
• Wide Neural Networks Forget Less Catastrophically
• Gradual Domain Adaptation in the Wild:When Intermediate Distributions are Absent
• Hyperplane Arrangements of Trained ConvNets Are Biased
• Efficient Meta Lifelong-Learning with Limited Memory
• Estimating informativeness of samples with Smooth Unique Information
• Investigating the Compositional Structure Of Deep Neural Networks
• An Orthogonal-SGD based Learning Approach for MIMO Detection under Multiple Channel Models
• Sampling Bias in Deep Active Classification: An Empirical Study
• SVP-CF: Selection via Proxy for Collaborative Filtering Data
• Reducing Geographic Performance Differential for Face Recognition
• Adaptive Discriminative Regularization for Visual Classification
• Data Summarization via Bilevel Optimization
• Multiplicative Reweighting for Robust Neural Network Optimization
• Learning and Memorizing Representative Prototypes for 3D Point Cloud Semantic and Instance Segmentation
• Accelerating Deep Learning with Dynamic Data Pruning
• Dynamic Knowledge Distillation for Pre-trained Language Models
• Mind Your Outliers! Investigating the Negative Impact of Outliers on Active Learning for Visual Question Answering
• Accelerating Neural Architecture Search via Proxy Data