arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 1910.02481cs.AI
收藏

通过神经逻辑归纳学习学习有效解释

Learn to Explain Efficiently via Neural Logic Inductive Learning

Georgia Institute of Technology(佐治亚理工学院)

Yuan Yang, Le Song

摘要:

做出可解释和自我解释的决策的能力对于开发负责任的机器学习系统至关重要。本文在归纳逻辑程序设计(ILP)的范围内研究了解释问题的学习。我们提出了神经逻辑归纳学习(NLIL),一个有效的可微ILP框架,学习一阶逻辑规则,可以解释数据中的模式。在实验中,与最先进的方法相比,我们发现NLIL可以搜索x10倍长的规则,同时保持x3倍的速度。我们还表明,NLIL可以扩展到大型图像数据集,即视觉基因组,1M实体。

英文摘要:

The capability of making interpretable and self-explanatory decisions is essential for developing responsible machine learning systems. In this work, we study the learning to explain problem in the scope of inductive logic programming (ILP). We propose Neural Logic Inductive Learning (NLIL), an efficient differentiable ILP framework that learns first-order logic rules that can explain the patterns in the data. In experiments, compared with the state-of-the-art methods, we find NLIL can search for rules that are x10 times longer while remaining x3 times faster. We also show that NLIL can scale to large image datasets, i.e. Visual Genome, with 1M entities.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑