XNNTab -- Interpretable Neural Networks for Tabular Data using Sparse Autoencoders

2025-12-15

AI Summary

To address the interpretability gap of neural networks in tabular data modeling—where their black-box nature hinders trust and regulatory compliance—this paper proposes a novel interpretable neural architecture that embeds Sparse Autoencoders (SAEs) into the backbone network. This integration enables semantically unambiguous decomposition of nonlinear features and explicit alignment with human-understandable concepts. The resulting model supports end-to-end interpretable prediction, with reasoning paths amenable to manual verification. Experiments demonstrate that our method matches the predictive accuracy of state-of-the-art black-box models (e.g., MLPs, XGBoost), while substantially outperforming conventional interpretable baselines (e.g., decision trees, linear models). The core contribution lies in the first deep structural coupling of SAEs as intrinsic interpretability modules within neural networks—thereby reconciling high representational capacity with concept-level transparency.

Technology Category

• Machine Learning: Transparent, Interpretable, Explainable ML
• Humans and AI: Explainable AI (XAI) for Human Understanding
• Natural Language Processing: Interpretability, Analysis, and Evaluation of NLP Models

Application Category

• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

In data-driven applications relying on tabular data, where interpretability is key, machine learning models such as decision trees and linear regression are applied. Although neural networks can provide higher predictive performance, they are not used because of their blackbox nature. In this work, we present XNNTab, a neural architecture that combines the expressiveness of neural networks and interpretability. XNNTab first learns highly non-linear feature representations, which are decomposed into monosemantic features using a sparse autoencoder (SAE). These features are then assigned human-interpretable concepts, making the overall model prediction intrinsically interpretable. XNNTab outperforms interpretable predictive models, and achieves comparable performance to its non-interpretable counterparts.

Problem

• Develops interpretable neural networks for tabular data
• Uses sparse autoencoders to decompose features into interpretable concepts
• Balances high predictive performance with model transparency

Innovation

• Uses sparse autoencoders for feature decomposition
• Assigns human-interpretable concepts to learned features
• Combines neural network expressiveness with interpretability