概率逻辑推理的弱相关分解方法

A WEAK-DEPENDENT-DECOMPOSITION APPROACH FOR PROBABILISTIC LOGIC

本文给出了一种新的概率逻辑推理的分解计算模型．若表示专家知识的概率逻辑公式集具有一定的弱相关分解特性，则可分解为多个公式子集．大规模线性规划的Ｄａｎｔｚｉｇ－Ｗｏｌｆｅ分解方法可用于对这种分解的问题求解．

This paper presents a new method for probabilistic logic inference. Ifthe logic formulas in a knowledge base keep certain weak-dependent-properties,they can be decomposed into several subsets. The Dantzig-Wolfe decomposition canbe used to solve such kind of problem.