On Baselines for Local Feature Attributions

Abstract

High-performing predictive models, such as neural nets, usually operate as black boxes, which raises serious concerns about their interpretability. Local feature attribution methods help to explain black box models and are therefore a powerful tool for assessing the reliability and fairness of predictions. To this end, most attribution models compare the importance of input features with a reference value, often called baseline. Recent studies show that the baseline can heavily impact the quality of feature attributions. Yet, we frequently find simplistic baselines, such as the zero vector, in practice. In this paper, we show empirically that baselines can significantly alter the discriminative power of feature attributions. We conduct our analysis on tabular data sets, thus complementing recent works on image data. Besides, we propose a new taxonomy of baseline methods. Our experimental study illustrates the sensitivity of popular attribution models to the baseline, thus laying the foundation for a more in-depth discussion on sensible baseline methods for tabular data.

Accepted at the AAAI-21 Workshop on Explainable Agency in AI

References in corpus (6)

• Learning Important Features Through Propagating Activation Differences
• SmoothGrad: removing noise by adding noise
• Sanity Checks for Saliency Maps
• Consistent Individualized Feature Attribution for Tree Ensembles
• Learning Parameter Distributions to Detect Concept Drift in Data Streams
• A Baseline for Shapley Values in MLPs: from Missingness to Neutrality

Cited by in corpus (5)

• Deep Neural Networks and Tabular Data: A Survey
• Fast Hierarchical Games for Image Explanations
• Choose Your Explanation: A Comparison of SHAP and GradCAM in Human Activity Recognition
• The Out-of-Distribution Problem in Explainability and Search Methods for Feature Importance Explanations
• A Robust Unsupervised Ensemble of Feature-Based Explanations using Restricted Boltzmann Machines