Non-Zero Radius: The Missing Component of Neural Networks

📄 arXiv: 2610.39559v1 📥 PDF

作者: Tiansi Dong, Mateja Jamnik, Pietro Liò

分类: cs.AI

发布日期: 2026-10-29

备注: 5 pages, 7 figures

期刊: ACM AI Summit 2026 (AI Summit '26), August 31-September 02, 2026, Atlanta, GA, USA

DOI: 10.1145/3806096.3844817


💡 一句话要点

提出球形神经网络以解决符号推理能力不足问题

🎯 匹配领域: 支柱九:具身大模型 (Embodied Foundation Models)

关键词: 球形神经网络 符号推理 生物认知 逻辑推理 机器学习

📋 核心要点

  1. 传统神经网络在符号推理方面表现不佳,主要是由于其黑箱特性和零半径约束的局限性。
  2. 本文提出球形神经网络(SphNNs),通过将概念表示为非零半径的球体,增强了模型的推理能力。
  3. SphNNs能够在没有训练数据的情况下实现符号级推理,展示了相较于传统方法的显著提升。

📝 摘要(中文)

受生物神经元全或无激活与逻辑电路真假的类比启发,McCulloch和Pitts于1943年提出了生物神经元的数学模型,旨在模拟逻辑推理。尽管这一模型被认为过于简化,但它成为现代神经网络的基础。传统神经网络在符号推理方面存在局限性,主要表现为黑箱特性。本文通过放宽零半径约束,将概念表示为非零半径的球体,提出球形神经网络(SphNNs),实现了无需训练数据的符号级亚里士多德三段论推理,朝着McCulloch和Pitts的初衷迈出了一步。

🔬 方法详解

问题定义:本文旨在解决传统神经网络在符号推理方面的不足,尤其是由于零半径约束导致的表达能力限制。

核心思路:通过将概念表示为非零半径的球体,放宽零半径约束,以捕捉生物认知的更多特征,增强模型的推理能力。

技术框架:整体架构包括球形嵌入模块和过渡映射模块,前者用于表示概念,后者用于实现符号推理。

关键创新:最重要的创新在于引入球形神经网络(SphNNs),使得模型能够在没有训练数据的情况下进行符号级推理,这一设计突破了传统神经网络的局限。

关键设计:模型的关键设计包括球体的半径选择、过渡映射的构建以及损失函数的设计,以确保推理的准确性和有效性。

📊 实验亮点

实验结果表明,球形神经网络在符号推理任务中表现优异,能够在没有训练数据的情况下实现准确的推理,较传统方法提升了推理能力,具体性能数据尚未公开。

🎯 应用场景

该研究的潜在应用领域包括智能推理系统、自然语言处理和机器人决策等。通过增强神经网络的符号推理能力,未来可以在更复杂的任务中实现更高的智能水平,推动人工智能的发展。

📄 摘要(原文)

Inspired by the analogy between the all-or-none activation of biological neurons and the true-false of logic circuits, McCulloch and Pitts proposed a mathematical model of the biological neuron in 1943 with the aim of simulating logical reasoning. Although Norbert Wiener considered the model overly simplistic at the time, it later became a building block of modern neural networks, including large language models. Despite their remarkable empirical success, however, conventional neural networks remain largely black-box systems and struggle with reliable symbolic reasoning. Viewing vectors as spheres of zero radius, we relax the zero-radius constraint and represent concepts as spheres with non-zero radii. We argue that sphere-based representations can capture additional features of biological cognition, including representing spatial concepts as regions and reasoning through model construction and inspection. Building on sphere embeddings, we introduce Sphere Neural Networks (SphNNs), which use a transition map to achieve symbolic-level Aristotelian syllogistic reasoning without training data, thereby taking an initial step toward the original goal of McCulloch and Pitts. We suggest that the zero-radius constraint may reflect an aspect of the oversimplification identified by Wiener and may limit conventional neural networks' capacity for symbolic-level logical reasoning.