欢迎您！东篱公司
退出
-
申报数据库
-
立项数据库
-
成果数据库
-
项目获奖数据库
中文摘要：
针对重叠社区的检测问题,描述非负矩阵因式分解方法和概率潜语义分析方法,说明这两种方法的等价性;基于矩阵因式分解方法,提出加权非负矩阵三因式分解方法,使用因式分解中的因子矩阵建立每个顶点的社区关系模型和社区之间的交互关系模型,为解决由于缺失数据产生的稀疏性问题,引入加权阵;在实际场景的数据集上进行实验,根据实验结果分析该方法对于自由参数的敏感度,验证了该方法性能优于一般非负矩阵因式分解方法。
英文摘要：
Aiming at the detection problem of overlapping communities,firstly,the nonnegative matrix factorization method and the probabilistic latent semantic analysis method were described,and the equivalence of these two methods was illustrated.Secondly,the weighted nonnegative matrix tri-factorization method was proposed based on the matrix factorization method.The community membership of each vertice and the interaction among communities were modeled using three factor matrices in the factorization.At the same time,weighted matrix was introduced to solve the sparsity problem due to missing data.Finally,some experiments on datasets in real scenarios were reported,coming to a conclusion that the proposed method is superior to general nonnegative matrix factorization method.And the sensitivity of this method for free parameters was analyzed.
同期刊论文项目
同项目期刊论文
A control structure selection method based on multivariable generalized predictive control for unsta
Interaction Analysis and Decomposition Principle for Control Structure Design of Large-scale Systems
期刊信息
位置：