Removing Spurious Concepts from Neural Network Representations via Joint Subspace Estimation
This addresses the issue of spurious correlations affecting model robustness for users in computer vision and NLP, representing an incremental improvement over prior concept-removal techniques.
The paper tackled the problem of out-of-distribution generalization in neural networks by removing spurious concepts without harming main-task features, and showed that their iterative algorithm outperformed existing methods on benchmark datasets like Waterbirds, CelebA, and MultiNLI.
Out-of-distribution generalization in neural networks is often hampered by spurious correlations. A common strategy is to mitigate this by removing spurious concepts from the neural network representation of the data. Existing concept-removal methods tend to be overzealous by inadvertently eliminating features associated with the main task of the model, thereby harming model performance. We propose an iterative algorithm that separates spurious from main-task concepts by jointly identifying two low-dimensional orthogonal subspaces in the neural network representation. We evaluate the algorithm on benchmark datasets for computer vision (Waterbirds, CelebA) and natural language processing (MultiNLI), and show that it outperforms existing concept removal methods