SGER: Fast Queries and Updates in Probabilistic Networks

卡西夫，西蒙·约翰·霍普金斯大学$30,620 -最多12美元。基于网络的知识表示是人工智能最有前途的方法之一。该研究人员目前获得的另一项资助的主要目标是为基于网络的人工智能工具（如约束网络和概率网络）开发非常高效的算法，重点是利用并行处理来实现网络规模的亚线性执行时间。这个现有项目的一个副产品是最近为概率网络提供了几种新的编译策略，这些策略允许在快速预处理之后，及时处理数据库大小的对数级查询和更新。这代表了当前算法的指数级改进。目前，已知这些技术适用于某些特殊情况，但似乎有可能将其扩展到更一般的网络。如果它们能够得到充分扩展，它将对技术产生强大的影响，因为定义、操作和更新概率信息的能力具有广泛的适用性。这笔用于探索性研究的小额赠款是为了探索这些技术的适用性，这些技术不属于现有赠款的研究计划的一部分。

IRI-9529227 Kasif, Simon Johns Hopkins University $30,620 - 12 mos. SGER: Fast Queries and Updates in Probabilistic Networks Knowledge representation using networks is one of the most promising approaches to Artificial Intelligence. The main goal of another current grant to the same investigator is to develop very efficient algorithms for network-based AI tools such as constraint networks and probabilistic networks, focusing on utilizing parallel processing to achieve execution time that is sub-linear in the size of the network. A by-product of this existing project recently was several new compilation strategies for probabilistic networks that allow, after fast preprocessing, the processing of both queries and updates in time that is logarithmic in the size of the database. This represents an exponential improvement in current algorithms. At present, the techniques are known to be applicable to certain special cases, but it appears possible to extend them to more general networks. If they can be extended adequately, it will have a strong impact on technology, given the wide applicability of the ability to define, manipulate, and update probabilistic information. This small grant for exploratory research is to explore the applicability of these techniques, which were not a part of the existing grant's research plan.