SaliencyDecor: Enhancing Neural Network Interpretability through Feature Decorrelation

2026-04-28

AI Summary

This work identifies feature correlation as a fundamental bottleneck in gradient-based saliency methods, which often yields ambiguous and unstable explanations that poorly align with semantically meaningful input regions. To address this, the authors propose a training framework that requires no modification to the model architecture. By jointly optimizing the classification loss, a prediction consistency constraint under feature masking, and a feature decorrelation regularizer, the method encourages the feature space to become approximately orthogonal during training, thereby concentrating gradient flow. Evaluated across multiple benchmark datasets and model architectures, the approach simultaneously improves both predictive accuracy and saliency map quality, producing sharper and more target-focused explanations.

Abstract

Gradient-based saliency methods are widely used to interpret deep neural networks, yet they often produce noisy and unstable explanations that poorly align with semantically meaningful input features. We argue that a fundamental cause of this behavior lies in the geometry of learned representations: correlated feature dimensions diffuse attribution gradients across redundant directions, resulting in blurred and unreliable saliency maps. To address this issue, we identify feature correlation as a structural limitation of gradient-based interpretability and propose SaliencyDecor, a training framework that enforces feature decorrelation to improve attribution fidelity without modifying saliency methods or model architectures by reshaping the feature space toward orthogonality, our approach promotes more concentrated gradient flow and improves the fidelity of saliency-based explanations. SaliencyDecor jointly optimizes classification, prediction consistency under feature masking, and a decorrelation regularizer, requiring no architectural changes or inference-time overhead. Extensive experiments across multiple benchmarks and architectures demonstrate that our method produces substantially sharper and more object-focused saliency maps while simultaneously improving predictive performance, achieving accuracy gains across the datasets. These results establish our method as a principled mechanism for enhancing both interpretability and accuracy, challenging the conventional trade-off between explanation quality and model performance.

Problem

• saliency
• feature correlation
• interpretability
• gradient-based explanation
• neural networks

Innovation

• feature decorrelation
• gradient-based saliency
• interpretability
• orthogonal feature space
• attribution fidelity