AutoHebb: Self-Modifying Neural Networks that Zero-Shot Infer and Continually Learn

查看原文 / View Original 编辑 / Edit

Computer Science > Machine Learning arXiv:2610.40757 (cs) [Submitted on 29 Oct 2026] Title:AutoHebb: Self-Modifying Neural Networks that Zero-Shot Infer and Continually Learn Authors:Daniel Levine, David van Dijk, Kenneth Kay View a PDF of the paper titled AutoHebb: Self-Modifying Neural Networks that Zero-Shot Infer and Continually Learn, by Daniel Levine and 2 other authors View PDF HTML (experimental)

计算机科学 > 机器学习 arXiv:2610.40757 (cs) [提交于 2026 年 10 月 29 日] 标题：AutoHebb：能够零样本推理和持续学习的自修改神经网络作者：Daniel Levine, David van Dijk, Kenneth Kay 查看标题为《AutoHebb：能够零样本推理和持续学习的自修改神经网络》的论文 PDF，作者为 Daniel Levine 及其他两位作者查看 PDF HTML（实验性）

Abstract:Humans and animals have a unique capacity to generalize and learn in novel circumstances, including “0-shot” inference (ZI) and continual learning (CL) — two powerful cognitive abilities. Despite this importance, efforts to model ZI and CL are seriously limited in several ways. First, existing models typically lack components critical to ZI and CL in the brain, namely, strong inductive biases (for ZI) and autonomous control of learning (for CL). Second, models can be less tractable to analysis if model components are overly complex (lacking simplicity), or too task-restricted if model components are overly bespoke (lacking expressivity).

摘要：人类和动物具有在陌生环境下进行归纳和学习的独特能力，包括“零样本”推理 (ZI) 和持续学习 (CL)——这两种强大的认知能力。尽管它们很重要，但模拟 ZI 和 CL 的努力在多个方面受到严重限制。首先，现有模型通常缺乏大脑中对于 ZI 和 CL 至关重要的组件，即强大的归纳偏置（针对 ZI）和自主学习控制（针对 CL）。其次，如果模型组件过于复杂（缺乏简洁性），则模型可能难以分析；如果模型组件过于定制化（缺乏表达能力），则模型可能过于受限于特定任务。

Here we introduce autonomous Hebbian networks (“AutoHebb”), a class of online learning feedforward neural networks that realize strong inductive biases and autonomously controlled learning, and are both simple and expressive. Unlike standard models that learn via externally controlled optimization (e.g. gradient descent), AutoHebb networks combine external feedback with internally generated activity to produce a global learning signal (‘neuromodulator’) fed back to the network. The neuromodulator in turn gates a biologically grounded local learning rule (3-factor Hebbian rule) that updates networks’ plastic weights.

在此，我们引入自主赫布网络（“AutoHebb”），这是一类实现在线学习的前馈神经网络，它们具备强大的归纳偏置和自主学习控制能力，且既简洁又具有表达力。与通过外部控制优化（如梯度下降）进行学习的标准模型不同，AutoHebb 网络将外部反馈与内部生成的活动相结合，产生一个反馈给网络的全局学习信号（“神经调节器”）。该神经调节器进而门控一种基于生物学的局部学习规则（三因子赫布规则），从而更新网络的塑性权重。

Using meta-learning to search for fixed (non-plastic) weights, we find that AutoHebb networks (1) succeed on ZI tasks, (2) show emergent (untrained) ZI/CL abilities, (3) admit reverse-engineering into explicit algorithms, and (4) outperform modern learning models on ZI/CL tasks. Lastly, performance comparisons to related models implied an essential role for three model components: meta-learning, Hebbian plasticity, and autonomous control of learning. These findings indicate that AutoHebb may be useful in hypothesis generation, analysis, and discovery of biologically grounded forms of learning and generalization.

通过使用元学习来搜索固定（非塑性）权重，我们发现 AutoHebb 网络 (1) 在 ZI 任务上取得成功，(2) 展示了涌现的（未经训练的）ZI/CL 能力，(3) 可以被逆向工程为显式算法，以及 (4) 在 ZI/CL 任务上优于现代学习模型。最后，与相关模型的性能比较表明，三个模型组件起到了至关重要的作用：元学习、赫布塑性和自主学习控制。这些发现表明，AutoHebb 可能在假设生成、分析以及发现基于生物学的学习和归纳形式方面具有应用价值。