Learning by Applying: A General Framework for Mathematical Reasoning via Enhancing Explicit Knowledge Learning

被引:0
|
作者
Liu, Jiayu [1 ,2 ,3 ]
Huang, Zhenya [1 ,2 ,3 ]
Zhai, Chengxiang [4 ]
Liu, Qi [1 ,2 ,3 ]
机构
[1] Univ Sci & Technol China, Sch Data Sci, Anhui Prov Key Lab Big Data Anal & Applicat, Hefei, Peoples R China
[2] Univ Sci & Technol China, Sch Comp Sci & Technol, Hefei, Peoples R China
[3] Univ Illinois, State Key Lab Cognit Intelligence, Champaign, IL USA
[4] Univ Illinois, Champaign, IL USA
来源
THIRTY-SEVENTH AAAI CONFERENCE ON ARTIFICIAL INTELLIGENCE, VOL 37 NO 4 | 2023年
基金
中国国家自然科学基金;
关键词
D O I
暂无
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Mathematical reasoning is one of the crucial abilities of general artificial intelligence, which requires machines to master mathematical logic and knowledge from solving problems. However, existing approaches are not transparent (thus not interpretable) in terms of what knowledge has been learned and applied in the reasoning process. In this paper, we propose a general Learning by Applying (LeAp) framework to enhance existing models (backbones) in a principled way by explicit knowledge learning. In LeAp, we perform knowledge learning in a novel problem-knowledge-expression paradigm, with a Knowledge Encoder to acquire knowledge from problem data and a Knowledge Decoder to apply knowledge for expression reasoning. The learned mathematical knowledge, including word-word relations and word-operator relations, forms an explicit knowledge graph, which bridges the knowledge "learning" and "applying" organically. Moreover, for problem solving, we design a semantics-enhanced module and a reasoning-enhanced module that apply knowledge to improve the problem comprehension and symbol reasoning abilities of any backbone, respectively. We theoretically prove the superiority of LeAp's autonomous learning mechanism. Experiments on three real-world datasets show that LeAp improves all backbones' performances, learns accurate knowledge, and achieves a more interpretable reasoning process.
引用
收藏
页码:4497 / 4506
页数:10
相关论文
共 50 条
  • [31] MemoryPath: A deep reinforcement learning framework for incorporating memory component into knowledge graph reasoning
    Li, Shuangyin
    Wang, Heng
    Pan, Rong
    Mao, Mingzhi
    NEUROCOMPUTING, 2021, 419 : 273 - 286
  • [32] A Meta-learning Knowledge Reasoning Framework Combining Semantic Path and Language Model
    Duan Li
    Feng Haojun
    Zhang Biying
    Liu Jiangzhou
    Liu Haichao
    JOURNAL OF ELECTRONICS & INFORMATION TECHNOLOGY, 2022, 44 (12) : 4376 - 4383
  • [34] Explicit General Analogy for Autonomous Transversal Learning
    Sheikhlar, Arash
    Thorisson, Kristinn R.
    Thompson, Jeff
    INTERNATIONAL WORKSHOP ON SELF-SUPERVISED LEARNING, VOL 192, 2022, 192 : 48 - 62
  • [35] The emergence of explicit knowledge from implicit learning
    Goujon, Annabelle
    Didierjean, Andre
    Poulet, Sarah
    MEMORY & COGNITION, 2014, 42 (02) : 225 - 236
  • [36] Implicit and explicit phonological knowledge in learning to read
    Gombert, JE
    INTERNATIONAL JOURNAL OF PSYCHOLOGY, 2000, 35 (3-4) : 90 - 90
  • [37] The emergence of explicit knowledge from implicit learning
    Annabelle Goujon
    André Didierjean
    Sarah Poulet
    Memory & Cognition, 2014, 42 : 225 - 236
  • [38] The Influence of Learning on Evolution: A Mathematical Framework
    Paenke, Ingo
    Kawecki, Tadeusz J.
    Sendhoff, Bernhard
    ARTIFICIAL LIFE, 2009, 15 (02) : 227 - 245
  • [39] A mathematical framework for cellular learning automata
    Beigy, H
    Meybodi, MR
    ADVANCES IN COMPLEX SYSTEMS, 2004, 7 (3-4): : 295 - 319
  • [40] Guiding Mathematical Reasoning via Mastering Commonsense Formula Knowledge
    Liu, Jiayu
    Huang, Zhenya
    Ma, Zhiyuan
    Liu, Qi
    Chen, Enhong
    Su, Tianhuang
    Liu, Haifeng
    PROCEEDINGS OF THE 29TH ACM SIGKDD CONFERENCE ON KNOWLEDGE DISCOVERY AND DATA MINING, KDD 2023, 2023, : 1477 - 1488