神经细胞自动机推理

中文摘要 AI 辅助

现代人工智能架构在解决视觉推理任务时，通常严重依赖全局连接和同步。然而，正如生物系统所示，复杂的计算可以以更分散的方式进行。在这项工作中，我们测试了神经细胞自动机（NCAs）的推理能力，这是一种使用严格局部连接和异步更新的循环细胞网络。NCAs 在人工生命实验中已被广泛研究，但尚不清楚它们是否能执行复杂的多步推理。我们表明，NCAs 产生的时空动力学能够解决具有挑战性的视觉推理任务，包括大型迷宫、数独和 ARC-AGI-1。此外，我们提供证据表明，当在更大的网格、更长的 rollout 或并行试验中运行时，NCAs 能够进行分布外泛化；并且通过修剪冗余轨迹，后者可以变得更高效。我们发现这些泛化能力依赖于样本重放和随机扰动的训练，并且随机性在测试时仍然有益。最后，我们表明 NCAs 是稳健的推理器，能够动态调节计算量以从损伤中高效恢复，并且它们可以扩展到在原始像素空间中解决推理问题。

英文摘要

Modern AI architectures used to solve visual reasoning tasks typically rely heavily on global connectivity and synchronization. As biological systems demonstrate, though, sophisticated computation can be performed in a more decentralized fashion. In this work, we test the reasoning capabilities of Neural Cellular Automata (NCAs), networks of recurrent cells that use strictly local connectivity and asynchronous updates. NCAs have been extensively studied in artificial life experiments, but it is unclear whether they can perform complex multi-step reasoning. We show that NCAs produce spatio-temporal dynamics capable of solving challenging visual reasoning tasks, including large mazes, Sudoku, and ARC-AGI-1. Furthermore, we provide evidence that NCAs generalize out-of-distribution when running with larger grids, longer rollouts, or parallel trials; and that the latter can be made more efficient via pruning of redundant trajectories. We find that these generalization capabilities depend on training with sample replay and stochastic perturbations, and that stochasticity remains beneficial at test time. Finally, we show that NCAs are robust reasoners capable of dynamically modulating compute to recover efficiently from damage, and that they can scale to solve reasoning in raw pixel space.

发表机构

• Google Paradigms of Intelligence Team（谷歌智能范式团队）
• School of Computer Science, McGill University（麦吉尔大学计算机科学学院）
• Mila - Quebec AI Institute（米拉-魁北克人工智能研究所）
• University of Geneva（日内瓦大学）
• Department of Neurology and Neurosurgery, McGill University（麦吉尔大学神经病学与神经外科系）
• Montreal Neurological Institute, McGill University（麦吉尔大学蒙特利尔神经研究所）
• Learning in Machines and Brains Program, CIFAR（CIFAR机器与大脑学习项目）

机构由 AI 辅助整理，请以论文原文为准。