LearningAsInferencePublic1
Part 2
LearningAsInferencePublic11
𝜃 is the collection of the probability of each state that we need to estimate from the data   𝑥 1 , …,  𝑥 𝑁
LearningAsInferencePublic11
 −  log 𝑝(𝑥|𝜃)   𝑞(𝑥) 
=− 𝑥  𝑞 𝑥  log 𝑝(𝑥|𝜃)  
=− 1 𝑁  𝑥   𝑛=1 𝑁 Ι[𝑥= 𝑥 𝑛 ]  log 𝑝(𝑥|𝜃)  
=− 1 𝑁  𝑛=1 𝑁  𝑥  Ι[𝑥= 𝑥 𝑛  ]  log 𝑝(𝑥|𝜃) 
=− 1 𝑁  𝑛=1 𝑁  log 𝑝( 𝑥 𝑛 |𝜃)
LearningAsInferencePublic12
LearningAsInferencePublic12
LearningAsInferencePublic13
LearningAsInferencePublic13
LearningAsInferencePublic9
LearningAsInferencePublic9
LearningAsInferencePublic10
LearningAsInferencePublic10
LearningAsInferencePublic10
LearningAsInferencePublic15
LearningAsInferencePublic15
LearningAsInferencePublic15
LearningAsInferencePublic16
LearningAsInferencePublic16
LearningAsInferencePublic16
LearningAsInferencePublic17
Global
Independence
Local
Independence
LearningAsInferencePublic18
For multivariate, a Dirichlet prior
Would be used. Details see Section 9.4.3.
LearningAsInferencePublic18
For multivariate, a Dirichlet prior
Would be used. Details see Section 9.4.3.
LearningAsInferencePublic18
For multivariate, a Dirichlet prior
Would be used. Details see Section 9.4.3.
LearningAsInferencePublic19
LearningAsInferencePublic19
LearningAsInferencePublic19
LearningAsInferencePublic20