Just Train Twice: Improving Group Robustness without Training Group Information

Abstract

Standard training via empirical risk minimization (ERM) can produce models that achieve high accuracy on average but low accuracy on certain groups, especially in the presence of spurious correlations between the input and label. Prior approaches that achieve high worst-group accuracy, like group distributionally robust optimization (group DRO) require expensive group annotations for each training point, whereas approaches that do not use such group annotations typically achieve unsatisfactory worst-group accuracy. In this paper, we propose a simple two-stage approach, JTT, that first trains a standard ERM model for several epochs, and then trains a second model that upweights the training examples that the first model misclassified. Intuitively, this upweights examples from groups on which standard ERM models perform poorly, leading to improved worst-group performance. Averaged over four image classification and natural language processing tasks with spurious correlations, JTT closes 75% of the gap in worst-group accuracy between standard ERM and group DRO, while only requiring group annotations on a small validation set in order to tune hyperparameters.

arXiv commentInternational Conference on Machine Learning (ICML), 2021

Cited by in corpus (13)

• WILDS: A Benchmark of in-the-Wild Distribution Shifts
• Detecting Shortcut Learning for Fair Medical AI using Shortcut Testing
• Counterfactual Learning on Graphs: A Survey
• Where Does My Model Underperform? A Human Evaluation of Slice Discovery Algorithms
• Environment Inference for Invariant Learning
• Robustness Challenges in Model Distillation and Pruning for Natural Language Understanding
• A Fine-Grained Analysis on Distribution Shift
• Focus on the Common Good: Group Distributional Robustness Follows
• MABR: Multilayer Adversarial Bias Removal Without Prior Bias Knowledge
• Which Samples Should be Learned First: Easy or Hard?
• Combating Unknown Bias with Effective Bias-Conflicting Scoring and Gradient Alignment
• Contrastive Clustering: Toward Unsupervised Bias Reduction for Emotion and Sentiment Classification
• Learning from Mistakes based on Class Weighting with Application to Neural Architecture Search