Explain Yourself, Briefly! Self-Explaining Neural Networks with Concise Sufficient Reasons

2025-02-05

AI Summary

This work addresses key challenges in generating minimal sufficient reasons—i.e., the smallest feature subsets guaranteeing invariant model predictions—for neural networks: high computational cost, reliance on out-of-distribution sampling, and post-hoc explanation distortion. We propose Self-Supervised Sufficient Subset Training (SST), a novel paradigm that embeds sufficient subset discovery directly into end-to-end model training. SST employs discrete subset parameterization and self-supervised optimization, enabling the network to natively output concise and faithful sufficient explanations at inference time. Compared to dominant post-hoc methods, SST eliminates iterative search, significantly improving explanation efficiency; enhances explanation faithfulness; and preserves original predictive accuracy. Its core innovation lies in the first end-to-end learnable and native generation of sufficient reasons—bypassing dependence on black-box predictors and mitigating distortions induced by out-of-distribution perturbations.

Technology Category

• Machine Learning: Unsupervised & Self-Supervised Learning
• Search and Optimization: Learning to Search
• Computer Vision: Adversarial Attacks & Robustness

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization

Abstract

Minimal sufficient reasons represent a prevalent form of explanation - the smallest subset of input features which, when held constant at their corresponding values, ensure that the prediction remains unchanged. Previous post-hoc methods attempt to obtain such explanations but face two main limitations: (1) Obtaining these subsets poses a computational challenge, leading most scalable methods to converge towards suboptimal, less meaningful subsets; (2) These methods heavily rely on sampling out-of-distribution input assignments, potentially resulting in counterintuitive behaviors. To tackle these limitations, we propose in this work a self-supervised training approach, which we term *sufficient subset training* (SST). Using SST, we train models to generate concise sufficient reasons for their predictions as an integral part of their output. Our results indicate that our framework produces succinct and faithful subsets substantially more efficiently than competing post-hoc methods, while maintaining comparable predictive performance.

Problem

• Generates concise sufficient reasons efficiently
• Addresses computational challenges in explanations
• Improves interpretability of neural networks

Innovation

• Self-supervised training approach
• Generates concise sufficient reasons
• Efficient than post-hoc methods