Neural Logic Machines
Honghua Dong 1, Jiayuan Mao1∗{}^{*}{}^{1}, Tian Lin2, Chong Wang3, Lihong Li2, and Denny Zhou2 1 ITCS, IIIS, Tsinghua University {dhh14, 2 Google Inc. 3 ByteDance Inc. indicates equal contribution. This work was done when the first two authors were interns at Google.
Abstract
We propose the Neural Logic Machine (NLM), a neural-symbolic architecture for both inductive learning and logic reasoning. NLMs exploit the power of both neural networks—as function approximators, and logic programming—as a symbolic processor for objects with properties, relations, logic connectives, and quantifiers. After being trained on small-scale tasks (such as sorting short arrays), NLMs can recover lifted rules, and generalize to large-scale tasks (such as sorting longer arrays). In our experiments, NLMs achieve perfect generalization in a number of tasks, from relational reasoning tasks on the family tree and general graphs, to decision making tasks including sorting arrays, finding shortest paths, and playing the blocks world. Most of these tasks are hard to accomplish for neural networks or inductive logic programming alone. 111Project page: https://sites.google.com/view/neural-logic-machines.
中文速览
神经逻辑机（Neural Logic Machine，NLM）要解决的核心难题是：纯神经网络不擅长逻辑推理和规则泛化，而传统归纳逻辑编程（ILP）又因搜索空间爆炸难以扩展到复杂任务。研究者提出了一种神经-符号混合架构，用张量表示逻辑谓词，用多层感知机模拟布尔逻辑运算（AND/OR/NOT），用扩展和归约操作实现全称/存在量词，从而把一阶逻辑的Horn子句"神经网络化"，让模型既能从数据中端到端学习，又能提取出可泛化的"提升规则"。实验表明，NLM在家谱关系推理、图上关系判断、数组排序、最短路径和积木世界等一系列任务上均达到完美泛化——尤其是在小规模样本（如短数组）上训练后，可直接推广到大规模实例（如长数组），而这些任务对单纯的神经网络或传统ILP来说都极具挑战性。这项工作的意义在于为神经网络引入了可扩展、可泛化的逻辑推理能力，朝着将深度学习与符号推理真正融合又迈出了重要一步。
原文 arXiv:1904.11694;中英对照 + 大白话阅读 https://aha.fim.ai/paper/1904.11694v1