0.25
0.5
0.75
1.25
1.5
1.75
2
Learning Scale Free Networks by Reweighted L1 regularization
Published on May 06, 20113942 Views
Methods for L1-type regularization have been widely used in Gaussian graphical model selection tasks to encourage sparse structures. However, often we would like to include more structural information
Related categories
Chapter list
Learning Scale-Free Networks by Reweighted L1 Regularization00:00
High Dimensional Structure Learning (1)00:16
High Dimensional Structure Learning (2)00:33
High Dimensional Structure Learning (3)00:35
Prior Information is important (1)01:08
Prior Information is important (2)02:01
Scale-Free Networks03:08
Barabási–Albert (B-A) model (1)04:26
Barabási–Albert (B-A) model (2)04:56
Barabási–Albert (B-A) model (3)05:16
Gaussian Markov Random Field05:36
Useful Properties of Gaussian06:54
L1 - based methods (1)08:04
L1 - based methods (2)09:01
L1 - based methods (3)09:41
L1 - based methods (4)10:03
L1 is not good for scale free networks11:15
Power Law Regularization (1)12:07
Power Law Regularization (2)12:56
L1 Relaxation (1)13:03
L1 Relaxation (2)13:16
L1 Relaxation (3)14:02
How to solve the optimization problem14:25
A MM algorithm (1)15:05
A MM algorithm (2)15:32
A MM algorithm (3)15:36
A MM algorithm (4)15:44
A MM algorithm (5)15:49
A MM algorithm (6)16:01
Reweighted L1 -based Optimization (1)16:24
Reweighted L1 -based Optimization (2)17:01
Reweighted L1 -based Optimization (3)17:34
Reweighted L1 -based Optimization (4)17:49
Reweighted L1 -based Optimization (5)18:16
Reweighted L1 -based Optimization (6)18:30
Other works on structured Regularization19:20
Experiments (simulated scale free network)20:39
ROC curves (1)21:21
ROC curves (2)21:45
ROC curves (3)22:15
ROC curves (4)22:22
Degree distributions22:26
Percentage of edges connecting to hubs23:24
Improvement over iterations (1)23:48
Improvement over iterations (2)24:00
Improvement over iterations (3)24:05
Improvement over iterations (4)24:09
Improvement over iterations (5)24:10
Experiments (simulated hub network)24:32
Experiments (Microarray data)24:59
Future works25:22
Thank you26:15