Class Unbiasing for Generalization in Medical Diagnosis
This addresses biased performance and poor generalization in medical diagnosis models, particularly for underperforming classes, but is incremental as it builds on existing bias mitigation techniques.
The paper tackles the problem of class-feature bias in medical diagnosis models, where reliance on features correlated with only some classes leads to poor generalization, and proposes a method that mitigates this bias and class imbalance, improving generalization as demonstrated on synthetic and real-world datasets.
Medical diagnosis might fail due to bias. In this work, we identified class-feature bias, which refers to models' potential reliance on features that are strongly correlated with only a subset of classes, leading to biased performance and poor generalization on other classes. We aim to train a class-unbiased model (Cls-unbias) that mitigates both class imbalance and class-feature bias simultaneously. Specifically, we propose a class-wise inequality loss which promotes equal contributions of classification loss from positive-class and negative-class samples. We propose to optimize a class-wise group distributionally robust optimization objective-a class-weighted training objective that upweights underperforming classes-to enhance the effectiveness of the inequality loss under class imbalance. Through synthetic and real-world datasets, we empirically demonstrate that class-feature bias can negatively impact model performance. Our proposed method effectively mitigates both class-feature bias and class imbalance, thereby improving the model's generalization ability.