Standard empirical risk minimization (ERM) training can produce deep neural network (DNN) models that are accurate on average but under-perform in under-represented population subgroups, especially when there are imbalanced group distributions in the long-tailed training data. Therefore, approaches that improve the accuracy-group robustness trade-off frontier of a DNN model (i.e. improving worst-group accuracy without sacrificing average accuracy, or vice versa) is of crucial importance. Uncertainty-based active learning (AL) can potentially improve the frontier by preferentially sampling underrepresented subgroups to create a more balanced training dataset. However, the quality of uncertainty estimates from modern DNNs tend to degrade in the presence of spurious correlations and dataset bias, compromising the effectiveness of AL for sampling tail groups. In this work, we propose Introspective Self-play (ISP), a simple approach to improve the uncertainty estimation of a deep neural network under dataset bias, by adding an auxiliary introspection task requiring a model to predict the bias for each data point in addition to the label. We show that ISP provably improves the bias-awareness of the model representation and the resulting uncertainty estimates. On two real-world tabular and language tasks, ISP serves as a simple "plug-in" for AL model training, consistently improving both the tail-group sampling rate and the final accuracy-fairness trade-off frontier of popular AL methods.
翻译:标准实证风险最小化(ERM)培训可产生深度神经网络(DNN)模型,这些模型平均准确,但在代表性不足的人口分组中表现不佳,特别是当长期培训数据存在不平衡的分组分布时。因此,提高DNN模型(即提高最差群体准确度而不牺牲平均准确性,或反之)准确性交易前沿的方法至关重要。基于不确定性的积极学习(AL)可能改善前沿,优先抽样代表性不足的分组,以创造更平衡的培训数据集。然而,现代DNNS的不确定性估计质量在存在虚假的关联和数据集偏差的情况下往往会下降,损害AL对抽样尾组的效益。在这项工作中,我们提议一种简单的方法,即改进基于数据集偏差的深层神经网络的不确定性估计,增加一个辅助性内向任务,要求为标签外的每个数据点预测偏差。我们表明,ISP的准确性估算质量,同时提高ILO的准确性,并持续改进ILA格式的准确性评估。