Tackling Shortcut Learning in Deep Neural Networks: An Iterative Approach with Interpretable Models
This addresses the problem of shortcut learning for AI practitioners by providing an interpretable method, though it appears incremental as it builds on existing interpretability techniques.
The paper tackles shortcut learning in deep neural networks by using concept-based interpretable models to detect and eliminate shortcuts without harming accuracy, achieving effective shortcut elimination as verified by First Order Logic explanations.
We use concept-based interpretable models to mitigate shortcut learning. Existing methods lack interpretability. Beginning with a Blackbox, we iteratively carve out a mixture of interpretable experts (MoIE) and a residual network. Each expert explains a subset of data using First Order Logic (FOL). While explaining a sample, the FOL from biased BB-derived MoIE detects the shortcut effectively. Finetuning the BB with Metadata Normalization (MDN) eliminates the shortcut. The FOLs from the finetuned-BB-derived MoIE verify the elimination of the shortcut. Our experiments show that MoIE does not hurt the accuracy of the original BB and eliminates shortcuts effectively.