1（School of Applied Mathematics,Guangdong University of Technology,Guangzhou 510520,China）2（School of Computer Science,Guangdong University of Technology,Guangzhou 510006,China）3（School of Mathematics and Big Data,Foshan University,Foshan 528000,China）
摘要 非高斯性数据间的因果网络已经在经济学、生物学和环境学等学科得到了广泛应用.DirectLingam（Direct Method for Learning a Linear Non-Gaussian Structural Equation Model）算法是其中一个经典解法,但其存在维度达到25维度以上时外生变量（exogenous variable）识别率低的问题,进而产生级联效应,使得整个网络的估计误差随着层数增大越来越大.为此提出了一种基于负熵局部选择外生变量的DirectLingam算法（LS-DirectLingam）,把变量的非高斯性作为外生变量选择的标准,用负熵来度量变量的非高斯,选择负熵最大的k个变量存入局部目标变量集合Lv中,在集合Lv中进一步去寻找外生变量,从而提高了外生变量的识别率.与基本的DirectLingam算法进行实验比较,结果表明LS-DirectLingam算法优于DirectLingam算法.
Abstract：Causal network between non-Gaussian data already in economics,biology,and environmental science and other disciplines has been widely used.DirectLingam (Direct Method for Learning a Linear Non-Gaussian Structural Equation Model) algorithm is one of the classic solution,but the problem of low recognition exogenous variables (exogenous variable) reaches a dimension more than 25 dimensions exist,and produces a cascading effect,that the entire network of the estimation error increases with increasing number of layers.To this end,we proposed DirectLingam algorithm based on negative entropy local choice exogenous variables (LS-DirectLingam),the non-Gaussian standard as an exogenous variable selection of variables with a negative entropy to measure non-Gaussian variables,selecting a negative entropy the largest k variables stored in a local variable target set in Lv,Lv in the collection further to find exogenous variables,thereby improving the recognition rate of exogenous variables.Compared with the basic DirectLingam algorithm,the results show that the LS-DirectLingam algorithm is better than the DirectLingam algorithm.
郝志峰(1,2,3),谢峰1,蔡瑞初2,陈薇2. 一种利用负熵局部选择外生变量的DirectLingam算法[J]. 小型微型计算机系统, 2017, 38(8): 1834-1839.
HAO Zhi-feng(1,2,3),XIE Feng1,CAI Rui-chu2,CHEN Wei2. DirectLingam Algorithm for Local Selection of Exogenous Variables by Using Negative Entropy. Journal of Chinese Computer Systems, 2017, 38(8): 1834-1839.
2017, Vol. 38 
