0.25
0.5
0.75
1.25
1.5
1.75
2
Learning Scale Free Networks by Reweighted L1 regularization
Published on 2011-05-063942 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
Presentation
Learning Scale-Free Networks by Reweighted L1 Regularization00:00
High Dimensional Structure Learning (1)04:42:04
High Dimensional Structure Learning (2)09:22:16
High Dimensional Structure Learning (3)09:48:20
Prior Information is important (1)19:04:07
Prior Information is important (2)33:37:01
Scale-Free Networks52:15:04
Barabási–Albert (B-A) model (1)74:08:37
Barabási–Albert (B-A) model (2)82:25:44
Barabási–Albert (B-A) model (3)87:59:21
Gaussian Markov Random Field93:24:28
Useful Properties of Gaussian115:15:50
L1 - based methods (1)134:27:35
L1 - based methods (2)150:27:36
L1 - based methods (3)161:25:21
L1 - based methods (4)167:45:33
L1 is not good for scale free networks187:31:49
Power Law Regularization (1)202:07:24
Power Law Regularization (2)215:38:58
L1 Relaxation (1)217:38:53
L1 Relaxation (2)221:14:35
L1 Relaxation (3)234:08:34
How to solve the optimization problem240:23:17
A MM algorithm (1)251:29:59
A MM algorithm (2)258:59:58
A MM algorithm (3)260:08:42
A MM algorithm (4)262:22:09
A MM algorithm (5)263:50:45
A MM algorithm (6)267:01:15
Reweighted L1 -based Optimization (1)273:28:49
Reweighted L1 -based Optimization (2)283:49:40
Reweighted L1 -based Optimization (3)292:57:02
Reweighted L1 -based Optimization (4)297:02:52
Reweighted L1 -based Optimization (5)304:40:11
Reweighted L1 -based Optimization (6)308:28:01
Other works on structured Regularization322:22:53
Experiments (simulated scale free network)344:17:34
ROC curves (1)355:54:59
ROC curves (2)362:35:30
ROC curves (3)370:55:49
ROC curves (4)372:48:01
Degree distributions373:57:48
Percentage of edges connecting to hubs390:03:53
Improvement over iterations (1)396:40:33
Improvement over iterations (2)400:11:50
Improvement over iterations (3)401:31:57
Improvement over iterations (4)402:45:17
Improvement over iterations (5)403:00:18
Experiments (simulated hub network)409:07:28
Experiments (Microarray data)416:38:03
Future works423:00:54
Thank you437:35:58