神经网络能否通过对自身进行实验来学习？用于预测自我知识的自干预学习
Can Neural Networks Learn by Experimenting on Themselves? Self-Interventional Learning from Functional Consequences to Predictive Self-Knowledge
浏览论文内容
中文总结 AI 辅助
本研究提出自干预学习（SIL）框架，让神经网络通过扰动自身功能结构学习预测自我模型，实验表明其可恢复部分网络属性，但自我模型存在不足且未显著优于简单策略。
中文摘要 AI 辅助
机器学习系统通常对外部数据进行建模，而其内部功能组织则由外部观察者分析。本研究引入自干预学习（Self-Interventional Learning, SIL），其中神经系统会扰动自身的功能结构、观测结果、学习预测自我模型、泛化到未执行的干预，并利用预测结果指导后续的结构动作。在一个结构已知的合成系统中，SIL 成功恢复了关键结构、冗余性和可替换性，但协同性未被可靠恢复。在 30 个全新的验证种子上，当成对干预预算从 4 增加到 56 时，保留样本预测误差从 0.0335 降至 0.0148，斯皮尔曼相关系数从 0.629 提升至 0.883。在匹配消融实验中，保留正确的干预-结果映射使前瞻性预测误差降低 81.3%，而使用相同的学习自我模型进行动作决策，相较于忽略该模型，归一化遗憾降低 31.7%。不过，模型指导的动作并未显著优于直接经验记忆策略，且在同等预算下，基于 CIFAR-10/ResNet 的验证显示其相比直接修复搜索无鲁棒性优势。这些结果表明，SIL 是一种基于干预的框架，用于学习关于网络自身功能组织的预测性知识，同时也显示该自我模型仍不完善，且并不普遍优于更简单的直接策略。
英文摘要
Machine-learning systems usually model external data, while their internal functional organization is analyzed by external observers. This work introduces Self-Interventional Learning (SIL), in which a neural system perturbs its own functional structure, observes consequences, learns a predictive self-model, generalizes to unexecuted interventions, and uses predictions to guide later structural action. In a construction-known synthetic system, SIL recovered critical structure, redundancy, and replaceability, while synergy was not reliably recovered. Across 30 fresh confirmatory seeds, increasing the pairwise intervention budget from 4 to 56 reduced held-out prediction error from 0.0335 to 0.0148 and increased Spearman correlation from 0.629 to 0.883. In a matched ablation, preserving the correct intervention--consequence mapping reduced prospective prediction error by 81.3%, while using the same learned self-model for action reduced normalized regret by 31.7% relative to ignoring it. However, model-guided action did not significantly outperform a direct empirical-memory policy, and powered CIFAR-10/ResNet validation showed no robustness advantage over equal-budget direct repair search. These results support SIL as an intervention-driven framework for learning predictive knowledge about a network's own functional organization, while showing that the self-model remains incomplete and is not universally superior to simpler direct strategies.
发表机构
- Opole University of Technology（奥波莱工业大学）
机构由 AI 辅助整理，请以论文原文为准。
