Input Bias in Rectified Gradients and Modified Saliency Maps
This addresses a critical flaw in interpretability methods for deep learning models, particularly for researchers and practitioners relying on saliency maps for model understanding, though it is incremental as it builds on existing techniques.
The paper identifies that Rectified Gradients and similar modified saliency maps introduce a strong input bias, such as favoring bright areas, which can misrepresent feature importance, and proposes a simple fix by removing multiplication with input features to eliminate this bias.
Interpretation and improvement of deep neural networks relies on better understanding of their underlying mechanisms. In particular, gradients of classes or concepts with respect to the input features (e.g., pixels in images) are often used as importance scores or estimators, which are visualized in saliency maps. Thus, a family of saliency methods provide an intuitive way to identify input features with substantial influences on classifications or latent concepts. Several modifications to conventional saliency maps, such as Rectified Gradients and Layer-wise Relevance Propagation (LRP), have been introduced to allegedly denoise and improve interpretability. While visually coherent in certain cases, Rectified Gradients and other modified saliency maps introduce a strong input bias (e.g., brightness in the RGB space) because of inappropriate uses of the input features. We demonstrate that dark areas of an input image are not highlighted by a saliency map using Rectified Gradients, even if it is relevant for the class or concept. Even in the scaled images, the input bias exists around an artificial point in color spectrum. Our modification, which simply eliminates multiplication with input features, removes this bias. This showcases how a visual criteria may not align with true explainability of deep learning models.