Learning Interpretable Logic Rules from Deep Vision Models

2025-03-13

AI Summary

Existing interpretability methods for deep vision models—such as saliency maps—suffer from fundamental limitations including causal incompleteness, overconfidence, and semantic ambiguity. To address these, we propose VisionLogic, the first framework enabling end-to-end translation of arbitrary vision models equipped with fully connected heads into causally verifiable formal logic rules. VisionLogic anchors semantics of final-layer neurons as visual predicates, validates their causal associations with real-world concepts, and induces both class-level global logic rules and instance-specific local explanations. Crucially, it requires no model retraining and preserves ≥95% of the original model’s discriminative performance. The generated logic rules are human-readable, formally verifiable, causally grounded, and support visual concept perturbation analysis—thereby substantially improving explanation reliability, model trustworthiness, and debuggability.

Technology Category

• Computer Vision: Large Vision Models
• Knowledge Representation and Reasoning: Logic Programming
• Machine Learning: Statistical Relational/Logic Learning

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Security and Privacy: Data transparency and provenance

Abstract

We propose a general framework called VisionLogic to extract interpretable logic rules from deep vision models, with a focus on image classification tasks. Given any deep vision model that uses a fully connected layer as the output head, VisionLogic transforms neurons in the last layer into predicates and grounds them into vision concepts using causal validation. In this way, VisionLogic can provide local explanations for single images and global explanations for specific classes in the form of logic rules. Compared to existing interpretable visualization tools such as saliency maps, VisionLogic addresses several key challenges, including the lack of causal explanations, overconfidence in visualizations, and ambiguity in interpretation. VisionLogic also facilitates the study of visual concepts encoded by predicates, particularly how they behave under perturbation -- an area that remains underexplored in the field of hidden semantics. Apart from providing better visual explanations and insights into the visual concepts learned by the model, we show that VisionLogic retains most of the neural network's discriminative power in an interpretable and transparent manner. We envision it as a bridge between complex model behavior and human-understandable explanations, providing trustworthy and actionable insights for real-world applications.

Problem

• Extracts interpretable logic rules from deep vision models
• Addresses lack of causal explanations in visualization tools
• Retains neural network's discriminative power transparently

Innovation

• Extracts logic rules from deep vision models
• Uses causal validation for grounding vision concepts
• Retains neural network's power interpretably