Towards the Unification and Robustness of Perturbation and Gradient Based Explanations

Abstract

As machine learning black boxes are increasingly being deployed in critical domains such as healthcare and criminal justice, there has been a growing emphasis on developing techniques for explaining these black boxes in a post hoc manner. In this work, we analyze two popular post hoc interpretation techniques: SmoothGrad which is a gradient based method, and a variant of LIME which is a perturbation based method. More specifically, we derive explicit closed form expressions for the explanations output by these two methods and show that they both converge to the same explanation in expectation, i.e., when the number of perturbed samples used by these methods is large. We then leverage this connection to establish other desirable properties, such as robustness, for these techniques. We also derive finite sample complexity bounds for the number of perturbations required for these methods to converge to their expected explanation. Finally, we empirically validate our theory using extensive experimentation on both synthetic and real world datasets.

arXiv commentThe short version of this paper appears in the proceedings of ICML-21

References in corpus (16)

• Axiomatic Attribution for Deep Networks
• Rényi Divergence and Kullback-Leibler Divergence
• Understanding Black-box Predictions via Influence Functions
• Interpretable classifiers using rules and Bayesian analysis: Building a better stroke prediction model
• SmoothGrad: removing noise by adding noise
• Certified Adversarial Robustness via Randomized Smoothing
• Sanity Checks for Saliency Maps
• Actionable Recourse in Linear Classification
• The Hidden Assumptions Behind Counterfactual Explanations and Principal Reasons
• Explanations can be manipulated and geometry is to blame
• Interpretability via Model Extraction
• Model-Agnostic Counterfactual Explanations for Consequential Decisions
• Algorithmic recourse under imperfect causal knowledge: a probabilistic approach
• Algorithmic Recourse: from Counterfactual Explanations to Interventions
• Certifiably Robust Interpretation in Deep Learning
• Classification by Set Cover: The Prototype Vector Machine