Greedy Structure Search for Sum-Product Networks

Greedy Structure Search for Sum-Product Networks
复制标题

和积网络的贪婪结构搜索

DOI:
--
复制
发表时间:
2015
期刊:
International Joint Conference on Artificial Intelligence
影响因子:
--
通讯作者:
D. Ventura
D. Ventura
中科院分区:
--
文献类型:
--
作者:
Aaron W. Dennis;D. Ventura

文献摘要

被引文献

相似文献

和积网络 (SPN) 是和与积节点的有根有向无环图 (DAG),具有明确定义的概率语义。此外,SPN 表示的分布的精确推断保证所用时间与 DAG 大小呈线性关系。在本文中,我们介绍了一种使用贪婪搜索方法学习 SPN 结构的算法。它结合了以前的 SPN 结构学习算法中使用的方法,但与以前的算法不同,它不限于学习树结构的 SPN。电路复杂性理论中的几个经过验证的想法以及我们的实验结果为具有较少限制的非树结构的 SPN 的优势提供了证据。
Sum-product networks (SPNs) are rooted, directed acyclic graphs (DAGs) of sum and product nodes with well-defined probabilistic semantics. Moreover, exact inference in the distribution represented by an SPN is guaranteed to take linear time in the size of the DAG. In this paper we introduce an algorithm that learns the structure of an SPN using a greedy search approach. It incorporates methods used in a previous SPN structure-learning algorithm, but, unlike the previous algorithm, is not limited to learning tree-structured SPNs. Several proven ideas from circuit complexity theory along with our experimental results provide evidence for the advantages of SPNs with less-restrictive, nontree structures.