A Tale Of Two Long Tails

Abstract

As machine learning models are increasingly employed to assist human decision-makers, it becomes critical to communicate the uncertainty associated with these model predictions. However, the majority of work on uncertainty has focused on traditional probabilistic or ranking approaches - where the model assigns low probabilities or scores to uncertain examples. While this captures what examples are challenging for the model, it does not capture the underlying source of the uncertainty. In this work, we seek to identify examples the model is uncertain about and characterize the source of said uncertainty. We explore the benefits of designing a targeted intervention - targeted data augmentation of the examples where the model is uncertain over the course of training. We investigate whether the rate of learning in the presence of additional information differs between atypical and noisy examples? Our results show that this is indeed the case, suggesting that well-designed interventions over the course of training can be an effective way to characterize and distinguish between different sources of uncertainty.

arXiv commentPreliminary results accepted to Workshop on Uncertainty and Robustness in Deep Learning (UDL), ICML, 2021

References in corpus (24)

• A Baseline for Detecting Misclassified and Out-of-Distribution Examples in Neural Networks
• Understanding deep learning requires rethinking generalization
• A Survey on Active Learning and Human-in-the-Loop Deep Learning for Medical Image Analysis
• Prototype selection for interpretable classification
• Fairness Without Demographics in Repeated Loss Minimization
• Online Batch Selection for Faster Training of Neural Networks
• Gradient Descent with Early Stopping is Provably Robust to Label Noise for Overparameterized Neural Networks
• The Bayesian Case Model: A Generative Approach for Case-Based Reasoning and Prototype Classification
• Coresets for Scalable Bayesian Logistic Regression
• Decomposition of Uncertainty in Bayesian Deep Learning for Efficient and Risk-sensitive Learning
• Pervasive Label Errors in Test Sets Destabilize Machine Learning Benchmarks
• What Do Compressed Deep Neural Networks Forget?
• Are we done with ImageNet?
• Revisiting the Calibration of Modern Neural Networks
• Probabilistic End-to-end Noise Correction for Learning with Noisy Labels
• Critical Learning Periods in Deep Neural Networks
• From ImageNet to Image Classification: Contextualizing Progress on Benchmarks
• Identifying Mislabeled Data using the Area Under the Margin Ranking
• Ordered SGD: A New Stochastic Optimization Framework for Empirical Risk Minimization
• Distribution Density, Tails, and Outliers in Machine Learning: Metrics and Applications
• A Study of Gradient Variance in Deep Learning
• Algorithmic Bias and Data Bias: Understanding the Relation between Distributionally Robust Optimization and Data Curation
• The Benchmark Lottery
• When does loss-based prioritization fail?