en-de
en-es
en-fr
en-sl
en
en-zh
0.25
0.5
0.75
1.25
1.5
1.75
2
The Discrete Infinite Logistic Normal Distribution for Mixed-Membership Modeling, incl. discussion by Frank Wood
Published on May 06, 20114283 Views
We present the discrete infinite logistic normal distribution (DILN, “Dylan”), a Bayesian nonparametric prior for mixed membership models. DILN is a generalization of the hierarchical Dirichlet pr
Related categories
Chapter list
The Discrete In nite Logistic Normal Distribution for Mixed-Membership Modeling00:00
Motivation (1)00:38
Motivation (2)01:04
Motivation (3)01:09
Motivation (4)01:19
Introduction: HDP (1)01:26
Introduction: HDP (2)02:44
Introduction: CTM (1)03:29
Introduction: CTM (2)04:51
Introduction: DILN (1)05:35
Introduction: DILN (2)05:54
Introduction: DILN (3)06:04
Introduction: DILN (4)06:26
Introduction: DILN (5)06:31
Review: The Dirichlet process and HDP (1)06:40
Review: The Dirichlet process and HDP (2)07:14
The discrete in nite logistic normal (1)08:48
The discrete in nite logistic normal (2)08:59
The discrete in nite logistic normal (3)09:14
The discrete in nite logistic normal (4)09:40
The discrete in nite logistic normal (5)09:59
Rough and Intuitive Example10:18
Constructing DILN (1)11:29
Constructing DILN (2)11:49
Constructing DILN (3)12:24
Constructing DILN (4)12:39
DILN Topic Model15:20
Variational inference for DILN (1)16:06
Variational inference for DILN (2)16:37
Variational inference for DILN (3)16:47
Variational inference for DILN (4)17:14
Variational inference for DILN (5)17:48
Experiments (1)19:06
Experiments (2)19:16
Experiments (3)19:22
Experiments (4)19:34
Experiments (5)19:46
Experiments (6)19:53
Experiments (7)19:58
Experiments: Perplexity Results20:08
Experiments: New York Times Topic Correlations20:41
Experiments: New York Times Topic Mapping21:28
Conclusion22:14
Future Work22:44
References22:59
Discussion of \The Discrete In nite Logistic Normal (DILN) Distribution for Mixed-Membership Modeling"23:10
Context (1)23:20
Context (2)23:30
Context (3)23:35
Context (4)23:40
Context (5)23:45
Context (6)23:50
Problem (1)24:12
Problem (2)24:35
Problem (3)25:12
Problem (4)25:25
DILN: Key Idea (1)25:32
DILN: Key Idea (2)25:35
DILN: Key Idea (3)25:45
DILN: Key Idea (4)25:55
DILN: Key Idea (5)26:07
Questions (1)26:17
Questions (2)26:42
Questions (3)27:05
Questions (4)27:15
Questions (5)27:20
Big Picture (1)27:42
Big Picture (2)27:48
Big Picture (3)27:54
References29:03