DifCluE: Generating Counterfactual Explanations with Diffusion Autoencoders and modal clustering

2025-02-17

AI Summary

Existing counterfactual explanation methods struggle to generate diverse, semantically meaningful counterfactuals for a single class, where multiple valid decision pathways exist. Method: We propose a within-class multimodal counterfactual generation framework based on a Diffusion Autoencoder (Diffusion AE). It integrates diffusion probabilistic models with variational autoencoders to explicitly model multimodal latent distributions; spectral clustering identifies distinct intra-class modal clusters, revealing interpretable sub-paths in the decision process; gradient-guided optimization then enables modality-controllable, manifold-aware counterfactual generation. Contribution/Results: Evaluated on multiple benchmark datasets, our method significantly outperforms state-of-the-art approaches. The generated counterfactuals exhibit high plausibility, diversity, and fidelity—demonstrating improved interpretability and trustworthiness of deep learning models without compromising predictive accuracy.

Technology Category

• Machine Learning: Deep Generative Models & Autoencoders
• Computer Vision: Diffusion Models for Vision
• Humans and AI: Explainable AI (XAI) for Human Understanding

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Economics, Online Markets and Human Computation: Economic ramifications for generative AI infrastructure and applications

Abstract

Generating multiple counterfactual explanations for different modes within a class presents a significant challenge, as these modes are distinct yet converge under the same classification. Diffusion probabilistic models (DPMs) have demonstrated a strong ability to capture the underlying modes of data distributions. In this paper, we harness the power of a Diffusion Autoencoder to generate multiple distinct counterfactual explanations. By clustering in the latent space, we uncover the directions corresponding to the different modes within a class, enabling the generation of diverse and meaningful counterfactuals. We introduce a novel methodology, DifCluE, which consistently identifies these modes and produces more reliable counterfactual explanations. Our experimental results demonstrate that DifCluE outperforms the current state-of-the-art in generating multiple counterfactual explanations, offering a significant advance- ment in model interpretability.

Problem

• Generates multiple counterfactual explanations
• Identifies distinct modes within a class
• Enhances model interpretability with DifCluE

Innovation

• Uses Diffusion Autoencoder for explanations
• Employs latent space clustering
• Generates diverse counterfactual explanations