HOLSTEP
收藏资源简介:
HOLSTEP数据集由因斯布鲁克大学开发,专注于高阶逻辑定理证明,包含2,013,046个训练样本和196,030个测试样本,源自11,400个证明。数据集涵盖基础数学、分析、三角学及图等数据结构的推理。创建过程涉及从HOL Light定理证明器中提取证明步骤,并进行离线处理以提取依赖关系和标记训练测试样本。该数据集旨在通过机器学习技术改进定理证明策略,特别是在交互式定理证明系统中提高自动化证明搜索的效率。
The HOLSTEP dataset, developed by the University of Innsbruck, focuses on higher-order logic theorem proving. It contains 2,013,046 training samples and 196,030 test samples, derived from 11,400 formal proofs. The dataset covers reasoning over data structures including foundational mathematics, analysis, trigonometry, and graphs. Its creation involves extracting proof steps from the HOL Light theorem prover, followed by offline processing to extract dependency relationships and label training and test samples. This dataset aims to improve theorem proving strategies via machine learning techniques, particularly to enhance the efficiency of automated proof search in interactive theorem proving systems.
- 1HolStep: A Machine Learning Dataset for Higher-order Logic Theorem Proving因斯布鲁克大学 · 2017年



