Evaluating the Robustness of Interpretability Methods through Explanation Invariance and Equivariance
This work addresses the need for robust interpretability methods in machine learning, particularly for users and developers dealing with models that have invariance properties, though it is incremental in building on geometric deep learning formalism.
The paper tackles the problem of ensuring interpretability methods faithfully describe models with symmetry groups, such as convolutional and graph neural networks, by formalizing explanation invariance and equivariance, deriving metrics and guidelines for robustness, and providing theoretical guarantees for some methods.
Interpretability methods are valuable only if their explanations faithfully describe the explained model. In this work, we consider neural networks whose predictions are invariant under a specific symmetry group. This includes popular architectures, ranging from convolutional to graph neural networks. Any explanation that faithfully explains this type of model needs to be in agreement with this invariance property. We formalize this intuition through the notion of explanation invariance and equivariance by leveraging the formalism from geometric deep learning. Through this rigorous formalism, we derive (1) two metrics to measure the robustness of any interpretability method with respect to the model symmetry group; (2) theoretical robustness guarantees for some popular interpretability methods and (3) a systematic approach to increase the invariance of any interpretability method with respect to a symmetry group. By empirically measuring our metrics for explanations of models associated with various modalities and symmetry groups, we derive a set of 5 guidelines to allow users and developers of interpretability methods to produce robust explanations.