arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2306.04019cs.AI
收藏

使用神经网络学习搜索空间特定的启发式函数

Learning Search-Space Specific Heuristics Using Neural Networks

University of Toronto(多伦多大学);The University of Tokyo(东京大学)

Yu Liu, Ryo Kuroiwa, Alex Fukunaga

AI总结:

本文提出一种系统,仅用单个PDDL实例并通过后向搜索生成数据,训练神经网络估计到目标距离,从而为经典规划学习搜索空间特定启发式函数,并在24-puzzle等领域展现出与领域无关启发式竞争的性能。

摘要:

我们提出并评估了一个系统,该系统为基于前向搜索的满足性经典规划学习神经网络启发式函数。给定单个PDDL训练实例,我们的系统从零开始学习到目标距离估计器。训练数据通过后向回归搜索,或从给定或猜测的目标状态进行后向搜索生成。在诸如24-puzzle这类所有实例共享同一搜索空间的领域中,此类启发式函数也可在该领域的所有实例间复用。我们表明,这个相对简单的系统能够取得令人惊讶的良好表现,有时可与著名的领域无关启发式函数相竞争。

英文摘要:

We propose and evaluate a system which learns a neuralnetwork heuristic function for forward search-based, satisficing classical planning. Our system learns distance-to-goal estimators from scratch, given a single PDDL training instance. Training data is generated by backward regression search or by backward search from given or guessed goal states. In domains such as the 24-puzzle where all instances share the same search space, such heuristics can also be reused across all instances in the domain. We show that this relatively simple system can perform surprisingly well, sometimes competitive with well-known domain-independent heuristics.

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。

↑