ml
10_graph_model
YeeKal
•
•
"#ml"
probabilistic graphical model
- vertex
- edge
- isomorphism [,aɪsə'mɔrfɪzm]
- directed graph
- undirected graph
- weight: weight/length/cost of edges
- graphical model
- bayesian network, 贝叶斯网络(directed acyclic graphical model(DAG),有向无环图`)
- markov random field, 马尔可夫随机场, undirected
bayesian network
贝叶斯网络中边代表连接关系,点代表在n个与之有连接关系的父节点同时出现的条件下的条件概率。
D-separation:
- head-to-head
- c未知的条件下,a,b独立
- tail-to-tail
- c已知的条件下,a,b独立
-
head-to-tail
- c已知的条件下,a,b独立
- 在较长的链式网络中,若$x_i$固定,则$x_{i+1}$与$x_i$之前的节点都独立。即$x_{i+1}$的状态只与前一个节点有关。这种顺次演变的随机过程,就叫做马尔科夫链(Markov chain).
markov network
MRF:马可夫无向图/马可夫随机场,markov random field
reference