Can Machine Learn Logics?

Can Machine Learn Logics?
复制标题

机器可以学习逻辑吗?

DOI:
10.1007/978-3-319-21365-1_35
复制
发表时间:
2015
期刊:
Artificial General Intelligence: Proceedings of the 8th International Conference , Lecture Notes in Artificial Intelligence
影响因子:
--
通讯作者:
Katsumi Inoue
Katsumi Inoue
中科院分区:
--
文献类型:
--
作者:
Chiaki Sakama;Katsumi Inoue

文献摘要

相似文献

本文讨论了设计能够从数据中学习逻辑的AI的可能性。我们提供了一个学习逻辑的抽象框架。在这个框架中,代理提供由公式及其逻辑结果组成的训练示例。然后,机器建立一个公理系统,作为Between和T之间的基础。或者,在没有代理的情况下,机器会寻找给定数据的未知逻辑。接下来,我们提供学习逻辑的两种情况:第一种情况考虑学习命题逻辑中的演绎推理规则,第二种情况考虑学习元胞自动机中的转换规则。每个案例研究都使用机器学习技术和元逻辑编程。
This paper argues the possibility of designing AI that can learn logics from data. We provide an abstract framework for learning logics. In this framework, an agentprovides training examples that consist of formulasSand their logical consequencesT. Then a machinebuilds an axiomatic system that underlies betweenSandT. Alternatively, in the absence of an agent, the machineseeks an unknown logic underlying given data. We next provide two cases of learning logics: the first case considers learning deductive inference rules in propositional logic, and the second case considers learning transition rules in cellular automata. Each case study uses machine learning techniques together with metalogic programming.