en
0.25
0.5
0.75
1.25
1.5
1.75
2
Mixed Cumulative Distribution Networks
Published on May 06, 20113414 Views
Directed acyclic graphs (DAGs) are a popular framework to express multivariate probability distributions. Acyclic directed mixed graphs (ADMGs) are generalizations of DAGs that can succinctly capture
Related categories
Chapter list
Mixed Cumulative Distribution Networks00:00
Directed Graphical Models00:14
Marginalization (1)00:57
Marginalization (2)01:24
The Acyclic Directed Mixed Graph (ADMG)01:55
Why do we care? (1)02:32
Why do we care? (2)03:36
The talk in a nutshell04:05
The Gaussian bi-directed model05:10
The Gaussian bi-directed case05:33
Binary bi-directed case: the constrained Moebius parameterization (1)06:06
Binary bi-directed case: the constrained Moebius parameterization (2)06:50
Binary bi-directed case: the constrained Moebius parameterization (3)07:42
The Cumulative Distribution Network (CDN) approach08:17
Relationship (1)09:15
Relationship (2)10:03
The Mixed CDN model (MCDN)10:54
Step 1: The high-level factorization (1)11:31
Step 1: The high-level factorization (2)12:30
Step 1: The high-level factorization (3)13:05
Step 2: Parameterizing P i (1)14:47
Step 2: Parameterizing P i (2)15:19
Step 2a: A copula formulation of P i (1)15:59
Step 2a: A copula formulation of P i (2)17:56
Step 2a: A copula formulation of P i (3)18:47
Step 2a: A copula formulation of P i (4)18:52
Step 2a: A copula formulation of P i (5)19:23
Step 3: The non-barren case (1)19:33
Step 3: The non-barren case (2)20:04
Step 3: The non-barren case (3)20:23
Parameter learning (1)20:29
Parameter learning (2)21:20
Parameter learning (3)21:30
Experiments (1)22:01
Experiments (2)22:31
Conclusion22:56
Acknowledgements23:59
Thank you24:06