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 2004.09855cs.AIcs.LG
收藏

通过超越蕴涵学习大型逻辑程序

Learning large logic programs by going beyond entailment

University of Oxford(牛津大学);KU Leuven(鲁汶大学)

Andrew Cropper, Sebastijan Dumančić

摘要:

归纳逻辑程序设计(ILP)的一个主要挑战是学习大型程序。我们认为,现有系统的一个关键限制是,他们使用蕴涵来指导假设搜索。这种方法是有限的,因为蕴涵是一个二元决策:一个假设要么需要一个例子,要么不需要,没有中间位置。为了解决这个问题,我们超越了蕴涵,并使用\emph{example-dependent}损失函数来指导搜索,其中假设可以部分覆盖一个例子。我们实现了我们的想法在Brute,一个新的ILP系统,它使用最好的优先搜索,由一个例子相关的损失函数的指导下,逐步建立程序。我们在三个不同的程序合成领域(机器人规划,字符串转换和ASCII艺术)的实验表明,Brute可以大大优于现有的ILP系统,无论是在预测精度和学习时间,可以学习的程序比最先进的系统大20倍。

英文摘要:

A major challenge in inductive logic programming (ILP) is learning large programs. We argue that a key limitation of existing systems is that they use entailment to guide the hypothesis search. This approach is limited because entailment is a binary decision: a hypothesis either entails an example or does not, and there is no intermediate position. To address this limitation, we go beyond entailment and use \emph{example-dependent} loss functions to guide the search, where a hypothesis can partially cover an example. We implement our idea in Brute, a new ILP system which uses best-first search, guided by an example-dependent loss function, to incrementally build programs. Our experiments on three diverse program synthesis domains (robot planning, string transformations, and ASCII art), show that Brute can substantially outperform existing ILP systems, both in terms of predictive accuracies and learning times, and can learn programs 20 times larger than state-of-the-art systems.

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。

↑