MUPAX: Multidimensional Problem Agnostic eXplainable AI

2025-07-17

AI Summary

Existing XAI methods struggle to simultaneously satisfy determinism, model-agnosticism, and theoretical convergence guarantees, while suffering from performance degradation under feature occlusion. This paper proposes MUPAX—a novel explainable AI framework grounded in measure-theoretic modeling and structured perturbation analysis. Its core innovation is the first formulation of multidimensional, task-agnostic feature importance attribution, endowed with rigorous mathematical convergence guarantees and full compatibility with arbitrary model architectures and loss functions. Empirically, MUPAX generates accurate, consistent, and semantically interpretable attributions across multimodal tasks—including audio classification, image recognition, 3D medical imaging, and anatomical keypoint detection—significantly outperforming state-of-the-art XAI baselines. Remarkably, under occlusion-based robustness testing, MUPAX not only preserves but *improves* model accuracy, effectively eliminating spurious correlations and uncovering intrinsic structural patterns in the data.

Technology Category

• Humans and AI: Explainable AI (XAI) for Human Understanding
• Machine Learning: Feature Construction/Reformulation
• Computer Vision: Interpretability, Explainability, and Transparency

Application Category

• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Web Mining and Content Analysis: Large pretrained models with web data
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Robust XAI techniques should ideally be simultaneously deterministic, model agnostic, and guaranteed to converge. We propose MULTIDIMENSIONAL PROBLEM AGNOSTIC EXPLAINABLE AI (MUPAX), a deterministic, model agnostic explainability technique, with guaranteed convergency. MUPAX measure theoretic formulation gives principled feature importance attribution through structured perturbation analysis that discovers inherent input patterns and eliminates spurious relationships. We evaluate MUPAX on an extensive range of data modalities and tasks: audio classification (1D), image classification (2D), volumetric medical image analysis (3D), and anatomical landmark detection, demonstrating dimension agnostic effectiveness. The rigorous convergence guarantees extend to any loss function and arbitrary dimensions, making MUPAX applicable to virtually any problem context for AI. By contrast with other XAI methods that typically decrease performance when masking, MUPAX not only preserves but actually enhances model accuracy by capturing only the most important patterns of the original data. Extensive benchmarking against the state of the XAI art demonstrates MUPAX ability to generate precise, consistent and understandable explanations, a crucial step towards explainable and trustworthy AI systems. The source code will be released upon publication.

Problem

• Develops deterministic, model-agnostic XAI with guaranteed convergence
• Provides principled feature attribution via structured perturbation analysis
• Enhances model accuracy by capturing key data patterns

Innovation

• Deterministic model agnostic XAI with convergence
• Feature importance via structured perturbation analysis
• Enhances accuracy by capturing key data patterns