DISCO: Mitigating Bias in Deep Learning with Conditional Distance Correlation
This work addresses bias mitigation for robust deep learning predictions, bridging causal theory with practical tools, though it appears incremental as it builds on existing bias mitigation approaches.
The paper tackles dataset bias in deep learning by introducing a causal framework and efficient estimators for conditional distance correlation, achieving competitive or superior performance across five datasets with fewer hyperparameters and scalability to multi-bias scenarios.
Dataset bias often leads deep learning models to exploit spurious correlations instead of task-relevant signals. We introduce the Standard Anti-Causal Model (SAM), a unifying causal framework that characterizes bias mechanisms and yields a conditional independence criterion for causal stability. Building on this theory, we propose DISCO$_m$ and sDISCO, efficient and scalable estimators of conditional distance correlation that enable independence regularization in black-box models. Across five diverse datasets, our methods consistently outperform or are competitive in existing bias mitigation approaches, while requiring fewer hyperparameters and scaling seamlessly to multi-bias scenarios. This work bridges causal theory and practical deep learning, providing both a principled foundation and effective tools for robust prediction. Source Code: https://github.com/***.