BayLIME: Bayesian Local Interpretable Model-Agnostic Explanations

Abstract

Given the pressing need for assuring algorithmic transparency, Explainable AI (XAI) has emerged as one of the key areas of AI research. In this paper, we develop a novel Bayesian extension to the LIME framework, one of the most widely used approaches in XAI -- which we call BayLIME. Compared to LIME, BayLIME exploits prior knowledge and Bayesian reasoning to improve both the consistency in repeated explanations of a single prediction and the robustness to kernel settings. BayLIME also exhibits better explanation fidelity than the state-of-the-art (LIME, SHAP and GradCAM) by its ability to integrate prior knowledge from, e.g., a variety of other XAI techniques, as well as verification and validation (V&V) methods. We demonstrate the desirable properties of BayLIME through both theoretical analysis and extensive experiments.

arXiv commentPreprint accepted by UAI2021. The final version to appear in the UAI2021 volume of Proceedings of Machine Learning Research

References in corpus (10)

• Grad-CAM++: Improved Visual Explanations for Deep Convolutional Networks
• Adversarial Examples Are Not Bugs, They Are Features
• RISE: Randomized Input Sampling for Explanation of Black-box Models
• Model Agnostic Supervised Local Explanations
• On the Robustness of Interpretability Methods
• Assessing Safety-Critical Systems from Operational Testing: A Study on Autonomous Vehicles
• Explaining Deep Learning Models - A Bayesian Non-parametric Approach
• What Do You See? Evaluation of Explainable Artificial Intelligence (XAI) Interpretability through Neural Backdoors
• Embedding and Extraction of Knowledge in Tree Ensemble Classifiers
• A Modified Perturbed Sampling Method for Local Interpretable Model-agnostic Explanation

Cited by in corpus (4)

• Reliable Post hoc Explanations: Modeling Uncertainty in Explainability
• Exploiting auto-encoders and segmentation methods for middle-level explanations of image classification systems
• Generative Perturbation Analysis for Probabilistic Black-Box Anomaly Attribution
• A Robust Unsupervised Ensemble of Feature-Based Explanations using Restricted Boltzmann Machines