This relation can convert the collapsibility of conditional Gaussian graphical models to corresponding problems on undirected graphs for considering. 这种关系能够帮助我们将条件高斯(CG)图模型的可压缩性问题转化到一般无向图上进行处理。
This tree can provide the conditional independence relations as much as possible for conditional Gaussian Bayesian networks. By using this tree, we can efficiently reduce the complexity of probability propagation computations. 这样得到的树能够尽可能多地体现出条件高斯(CG)贝叶斯网上的整体上的条件独立关系,从而,利用这棵树进行概率传播计算,能有效降低计算量和复杂度。
Deducing the natural gradient calculation formula for discrete BN, continuous BN, conditional Gaussian network and BN with discrete children of continuous parents. 推导出了离散型、连续型、条件Gaussian网、父节点连续而子节点离散等不同的贝叶斯网络类型其自然梯度的计算公式。
Effects of Correlative Structure of Rough Surface on the Conditional statistical Distributions of Phase Difference for Gaussian Speckle in Far-Field 粗糙表面相关结构对远场散斑相位差统计性质的影响
The relations between the surface roughness and the conditional statistical distributions of phase difference for far-field Gaussian speckle produced by a weak diffuser are investigated on condition that the height variation of rough surface is a Gaussian random variable and the rough surface obeys circle correlation. 在粗糙表面高度起伏服从高斯统计和表面服从圆型相关的情况下,分析了由弱散射体产生的远场高斯激光散斑相位差的条件统计分布与表面粗糙度的关系。