- 博客(3)
- 收藏
- 关注
原创 表示因果的三种图:DAG, MAG and PAG
表示因果的三种图:DAG, MAG and PAG1. DAG2. MAG(maximal ancestral graph)2.1 Mixed Graphs2.2 Ancestral Graphs2.3 Maximal Ancestral Graphsreferences1. DAGDirected acyclic graph,最常见的一种图。一般来说,A→BA\rightarrow BA→B 表示 A是B的原因。缺陷:1)在一个带有潜在变量的因果DAG下,干预后的概率是不可识别的;2)通过观测数
2021-04-26 21:28:03
3661
1
原创 Q&A about Transformer
文章目录What does the feed-forward neural network in Transformer architecture actually learn?What does the feed-forward neural network in Transformer architecture actually learn?A1: Actually, as far as I understand, the Self Attention layer doesn’t learn rep
2021-03-19 10:46:27
147
原创 Conditional Independent & D-seperation
Conditional Independent & d-seperation1. Three typical DAGs for conditional independence2. Conditional Independent2.1 Chain1. Three typical DAGs for conditional independence2. Conditional Independent2.1 ChainP(x,z,y)=P(x)P(z∣x)P(y∣z)⇒ConditionzP(x
2021-01-04 20:38:54
298
1
空空如也
空空如也
TA创建的收藏夹 TA关注的收藏夹
TA关注的人