Peano: learning formal mathematical reasoning.

Peano: learning formal mathematical reasoning.
复制标题

DOI:
10.1098/rsta.2022.0044
复制
发表时间:
2023-07-24
影响因子:
5
通讯作者:
Goodman, Noah D.
Goodman, Noah D.
中科院分区:
综合性期刊2区
文献类型:
--
作者:
Poesia, Gabriel;Goodman, Noah D.

文献摘要

参考文献

被引文献

相似文献

一般的数学推理在计算上是不可确定的,但人类通常会解决新问题。此外,几个世纪以来的发现很快就会传授给后代。是什么结构实现了这一点,它又如何为自动数学推理提供信息?我们假设这两个谜题的核心是数学基础的程序抽象结构。我们在可汗学院平台上的五个部分的初级代数的案例研究中探讨了这个想法。为了定义计算基础,我们引入了Peano,这是一个定理证明环境,其中任意点上的有效操作集是有限的。我们使用Peano来形式化入门代数问题和公理,得到定义良好的搜索问题。我们观察到现有的用于符号推理的强化学习方法不足以解决更难的问题。添加从其自己的解决方案中诱导可重用抽象(“策略”)的能力允许代理稳步前进,解决所有问题。此外,这些抽象归纳出在训练过程中随机出现的问题的顺序。恢复的秩序与专家设计的可汗学院课程有很大的一致性,在恢复的课程上训练的第二代代理学习速度要快得多。这些结果说明了抽象和课程在数学文化传播中的协同作用。本文是“认知人工智能”讨论会议的一部分。
General mathematical reasoning is computationally undecidable, but humans routinely solve new problems. Moreover, discoveries developed over centuries are taught to subsequent generations quickly. What structure enables this, and how might that inform automated mathematical reasoning? We posit that central to both puzzles is the structure of procedural abstractions underlying mathematics. We explore this idea in a case study on five sections of beginning algebra on the Khan Academy platform. To define a computational foundation, we introduce Peano, a theorem-proving environment where the set of valid actions at any point is finite. We use Peano to formalize introductory algebra problems and axioms, obtaining well-defined search problems. We observe existing reinforcement learning methods for symbolic reasoning to be insufficient to solve harder problems. Adding the ability to induce reusable abstractions (‘tactics’) from its own solutions allows an agent to make steady progress, solving all problems. Furthermore, these abstractions induce an order to the problems, seen at random during training. The recovered order has significant agreement with the expert-designed Khan Academy curriculum, and second-generation agents trained on the recovered curriculum learn significantly faster. These results illustrate the synergistic role of abstractions and curricula in the cultural transmission of mathematics. This article is part of a discussion meeting issue ‘Cognitive artificial intelligence’.
DOI: 10.1017/fmp.2017.1
发表时间: 2017-05-29
影响因子: 2.3
作者:
Hales, Thomas;Adams, Mark;Zumkeller, Roland
通讯作者: Zumkeller, Roland
DOI: 10.1006/ijhc.2000.0394
发表时间: 2000-09-01
影响因子: 5.4
作者:
Colton, S;Bundy, A;Walsh, T
通讯作者: Walsh, T
DOI: 10.1145/368273.368557
发表时间: 1962-01-01
影响因子: 22.7
作者:
DAVIS, M;LOGEMANN, G;LOVELAND, D
通讯作者: LOVELAND, D
DOI: 10.1145/321033.321034
发表时间: 1960-01-01
期刊: JOURNAL OF THE ACM
影响因子: 2.5
作者:
DAVIS, M;PUTNAM, H
通讯作者: PUTNAM, H
DOI: 10.1145/3065386
发表时间: 2017-06-01
影响因子: 22.7
作者:
Krizhevsky, Alex;Sutskever, Ilya;Hinton, Geoffrey E.
通讯作者: Hinton, Geoffrey E.