0.25
0.5
0.75
1.25
1.5
1.75
2
The Graph-guided Group Lasso
Published on 2013-08-264040 Views
In this work we propose a penalised regression model in which the covariates are known to be clustered into groups, and the clusters are arranged as nodes in a graph. We are motivated by an applicatio
Presentation
The Graph-guided Group Lasso00:00
Outline - 155:00
A 30s introduction to the biology16:18:20
Single-nucleotide polymorphisms (SNPs)20:08:20
Genome-wide association study (GWAs)24:49:22
Notation28:07:08
Sparse solution35:38:20
Penalized linear regression - 141:55:40
Penalized linear regression - 244:04:20
Some notable penalties that impose sparsity47:52:20
Incorporating prior biological knowledge - Variable grouping - 148:41:40
Incorporating prior biological knowledge - Variable grouping - 253:16:08
Incorporating prior biological knowledge - Variable grouping - 354:55:50
Incorporating prior biological knowledge - Variable grouping - 457:25:29
Incorporating prior biological knowledge - Variable grouping - 562:14:28
Incorporating prior biological knowledge - Network - 166:53:29
Incorporating prior biological knowledge - Network - 271:48:08
Incorporating prior biological knowledge - Network - 376:53:40
Incorporating prior biological knowledge - Network - 478:32:20
Incorporating prior biological knowledge - Network - 579:44:20
Incorporating prior knowledge at multiple levels93:26:57
The between-group relations103:32:21
Notation111:29:48
GGGL-1: Illustration125:09:36
GGGL-1: The model133:11:13
GGGL-1: Smoothing e ffect155:08:02
GGGL-1: A potential side e ffect176:30:42
GGGL-2: Another interpretation195:58:42
GGGL-2: The model - 1202:57:25
GGGL-2: The model - 2208:07:00
GGGL-2: Smoothing e ffect213:50:15
GGGL-2: Within-group eff ect224:24:53
Comparison: GGGL-1 and GGGL-2 smoothing e ffect - 1229:30:20
Comparison: GGGL-1 and GGGL-2 smoothing e ffect - 2232:17:40
Data generation: key settings - 1236:47:00
Data generation: key settings - 2242:47:00
Comparison: small for GGGL-1265:48:43
Comparison: large for GGGL-1275:51:23
Comparison: small for GGGL-2287:52:43
Comparison: large for GGGL-2298:27:40
Estimation algorithm: GGGL-1 - 1302:01:00
Estimation algorithm: GGGL-1 - 2305:55:40
Estimation algorithm: GGGL-1 - 3307:41:40
Estimation algorithm: GGGL-1 - 4310:29:39
Estimation algorithm: GGGL-1 - 5312:25:39
Estimation algorithm: GGGL-1 - 6317:48:49
Estimation algorithm: GGGL-2 - 1321:47:40
Estimation algorithm: GGGL-2 - 2325:20:20
Parallel computation: outline - 1329:03:40
Parallel computation: outline - 2341:04:36
Preliminary results364:54:00
Networks for GGGL374:01:40
Illustration of networks397:29:22
Experiment design: GGGL-1 vs Group lasso413:36:04
GGGL-1 vs Group lasso414:30:20
Future works419:08:20
Acknowledgement437:01:00
Reference441:50:20