Taxonomic Networks: A Representation for Neuro-Symbolic Pairing

2025-05-30

AI Summary

The disconnection between neural and symbolic approaches leads to inefficient and inflexible concept learning for hierarchical classification. Method: We propose a novel “neuro-symbolic pair” paradigm, built upon a bidirectional translatable framework anchored by a hierarchical classification network: the symbolic component models taxonomy structure efficiently using logical rules and few-shot examples; the neural component achieves high-accuracy discrimination via structured neural architectures; both sides are aligned within a unified knowledge representation space and support on-demand switching. Contribution/Results: We formally define the neuro-symbolic pair for the first time, enabling deep coupling at the knowledge representation level—not merely shallow integration. Experiments demonstrate that our framework simultaneously ensures robustness under data- and compute-constrained settings and attains competitive performance with abundant resources, significantly enhancing flexibility and generalization across diverse classification scenarios.

Technology Category

• Machine Learning: Neuro-Symbolic Learning
• Cognitive Modeling & Cognitive Systems: Symbolic Representations
• Computer Vision: Visual Reasoning & Symbolic Representations

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

We introduce the concept of a extbf{neuro-symbolic pair} -- neural and symbolic approaches that are linked through a common knowledge representation. Next, we present extbf{taxonomic networks}, a type of discrimination network in which nodes represent hierarchically organized taxonomic concepts. Using this representation, we construct a novel neuro-symbolic pair and evaluate its performance. We show that our symbolic method learns taxonomic nets more efficiently with less data and compute, while the neural method finds higher-accuracy taxonomic nets when provided with greater resources. As a neuro-symbolic pair, these approaches can be used interchangeably based on situational needs, with seamless translation between them when necessary. This work lays the foundation for future systems that more fundamentally integrate neural and symbolic computation.

Problem

• Develop neuro-symbolic pairs using common knowledge representation
• Introduce taxonomic networks for hierarchical concept organization
• Compare efficiency and accuracy of symbolic vs neural methods

Innovation

• Neuro-symbolic pairs linked by common knowledge
• Taxonomic networks organize concepts hierarchically
• Seamless translation between neural and symbolic methods