Abstract Interpretation-Based Feature Importance for SVMs

TL;DR

This paper introduces a novel abstract interpretation-based method for interpreting SVMs by deriving a dataset-independent feature importance measure and verifying stability and fairness, demonstrated to outperform existing importance measures.

Contribution

The paper presents a new abstract interpretation approach to improve SVM interpretability and stability verification, introducing the AFI measure that is dataset-independent and computationally efficient.

Findings

AFI correlates strongly with SVM stability to feature perturbations

AFI provides better insight into SVM trustworthiness than permutation importance

Method effective on linear and non-linear SVM kernels

Abstract

We propose a symbolic representation for support vector machines (SVMs) by means of abstract interpretation, a well-known and successful technique for designing and implementing static program analyses. We leverage this abstraction in two ways: (1) to enhance the interpretability of SVMs by deriving a novel feature importance measure, called abstract feature importance (AFI), that does not depend in any way on a given dataset of the accuracy of the SVM and is very fast to compute, and (2) for verifying stability, notably individual fairness, of SVMs and producing concrete counterexamples when the verification fails. We implemented our approach and we empirically demonstrated its effectiveness on SVMs based on linear and non-linear (polynomial and radial basis function) kernels. Our experimental results show that, independently of the accuracy of the SVM, our AFI measure correlates much more strongly with the stability of the SVM to feature perturbations than feature importance measures widely available in machine learning software such as permutation feature importance. It thus gives better insight into the trustworthiness of SVMs.