# Tabular diffusion counterfactual explanations

## TL;DR

This paper introduces a new method for generating counterfactual explanations using diffusion models for tabular data in fields like finance and social sciences.

## Contribution

A novel guided reverse process for categorical features using an approximation to the Gumbel-softmax distribution is proposed.

## Key findings

- The proposed method outperforms baseline methods in generating counterfactual explanations.
- The approach produces robust and realistic explanations on credit lending and other tabular datasets.
- A theoretical bound between the Gumbel-softmax and the proposed distribution is derived.

## Abstract

Counterfactual explanations methods provide an important tool in the field of interpretable machine learning. Recent advances in this direction have focused on diffusion models to explain a deep classifier. However, these techniques have predominantly focused on problems in computer vision. In this study, we focus on tabular data typical in finance and the social sciences and propose a novel guided reverse process for categorical features based on an approximation to the Gumbel-softmax distribution. Furthermore, we study the effect of the temperature τ and derive a theoretical bound between the Gumbel-softmax distribution and our proposed approximated distribution. We perform experiments on several large-scale credit lending and other tabular datasets, assessing their performance in terms of the quantitative measures of interpretability, diversity, instability, and validity. These results indicate that our approach outperforms popular baseline methods, producing robust and realistic counterfactual explanations.