A Bayesian network is:
A directed acyclic graph(DAG) G whose nodes represent the random variables X1,…,Xn
For each node Xi a CPD P(Xi | ParG(Xi))
The BN represents a joint distribution via the chain rule for Bayesian networks
P(X1,…,Xn) = Πi P(Xi | ParG(Xi))
BN is a legal distribution: P ≥ 0
BN is a legal distribution: ∑P = 1
∑X1,…,Xn P(X1,…,Xn) = ∑X1,…,Xn Πi P(Xi | ParG(Xi))
=∑X1,…,Xn-1 Πi P(Xi | ParG(Xi)) ∑Xn P(Xn | ParG(Xn)) //feel uncertain about this convert
=∑X1,…,Xn-1 Πi P(Xi | ParG(Xi)) = … … = 1 //green-part expression equal to 1
Causal Reasoning:
因果推理,由贝叶斯网络条件因素先验概率推导受影响因素的基本概率模型,是一个正向推理的过程(由条件分析结果)——先验推导
例如:基于I和D的先验概率推导 P(L|I,D)
Evidential Reasoning:
证据推理,由受影响因素的信息来推导条件因素的后验概率,是一个逆推的过程(由结果分析原因)——后验推导
例如:基于G的概率分布推导 P(I|G) 和 P(D|G)
Intercausal Reasoning:
因果联合推理,是一种混合推导
例如:基于现有的D和G的分布推导 P(I|D,G)
Flow of Probabilistic Influence(概率影响流):
Active Trails:A trail X1 ─ … ─ Xn is active if: it has no v-structures Xi-1 ~ Xi ~ Xi+1 (有效路径)
Evidence Trails(带有先验信息的有效路径):
上述图内描述可能有误,其中W应该是x和y之间的子路径,那么应该是evidence Z是属于W的形式描述,不过不影响理解
PPT以及其他资源来自于 coursera 上的 probability graphic model课程。
一些标注自己添加,文中可能理解不确切或者错误,欢迎交流。