Sign inSign up
PAC-Bayes Analysis of Classification thumbnail
Pause
Mute
Subtitles
Playback speed
0.25
0.5
0.75
1
1.25
1.5
1.75
2
Full screen

PAC-Bayes Analysis of Classification

Published on Dec 14, 20079970 Views

John Shawe-Taylor

The lecture will introduce the PAC Bayes approach to the statistical analysis of learning. After some historical introduction, the key theorems will be covered. We will then consider some applications

The Analysis of Patterns 2007 - Bertinoro

Related categories

Bayesian LearningSupport Vector Machines

Chapter list

PAC - Bayes Analysis of Classification00:00
Aims - 101:34
Aims - 201:35
Aims - 301:39
Aims - 401:40
Aims - 502:07
Aims - 602:09
Outline02:42
General perspectives - 103:48
General perspectives - 205:06
General perspectives - 305:19
General perspectives - 405:48
General perspectives - 506:07
Historical notes: Frequentist approach - 106:29
Historical notes: Frequentist approach - 206:52
Historical notes: Frequentist approach - 307:04
Historical notes: Frequentist approach - 408:30
Historical notes: Frequentist approach - 510:01
Historical notes: Bayesian approach - 110:08
Historical notes: Bayesian approach - 211:41
Historical notes: Bayesian approach - 313:04
Historical notes: Bayesian approach - 415:28
Version space: Evidence16:27
Evidence and generalisation - 119:01
Evidence and generalisation - 219:59
Evidence and generalisation - 320:34
Evidence and generalisation - 421:08
PAC-Bayes theorem - 121:40
PAC-Bayes theorem - 222:10
PAC-Bayes theorem - 322:31
PAC-Bayes theorem - 422:56
Definitions for main result: Prior and posterior distributions - 123:11
Definitions for main result: Prior and posterior distributions - 224:19
Definitions for main result: Prior and posterior distributions - 325:14
Definitions for main result: Error measures - 127:20
Definitions for main result: Error measures - 228:15
Definitions for main result: Error measures - 328:24
Definitions for main result: Error measures - 428:56
Definitions for main result: Assessing the posterior - 129:23
Definitions for main result: Assessing the posterior - 234:02
Definitions for main result: Generalisation error34:45
PAC-Bayes theorem - 536:10
Ingredients of proof (1/3) - 146:02
Ingredients of proof (1/3) - 247:20
Ingredients of proof (2/3) - 149:15
Ingredients of proof (2/3) - 249:40
Ingredients of proof (3/3) - 151:00
Ingredients of proof (2/3) - 351:29
Ingredients of proof (3/3) - 252:09
Finite classes - 152:54
Finite classes - 253:37
Other extensions/applications - 153:58
Other extensions/applications - 254:13
Other extensions/applications - 354:50
Other extensions/applications - 455:05
Linear classifiers and SVMs - 155:16
Linear classifiers and SVMs - 255:24
Linear classifiers and SVMs - 355:28
Linear classifiers and SVMs - 455:33
Linear classifiers - 156:26
Linear classifiers - 256:38
Linear classifiers - 356:41
PAC-Bayes bound for SVM (1/2) - 157:03
PAC-Bayes bound for SVM (1/2) - 257:09
PAC-Bayes bound for SVM (1/2) - 357:23
PAC-Bayes bound for SVM (1/2) - 457:26
PAC-Bayes bound for SVM (2/2) - 157:35
PAC-Bayes bound for SVM (2/2) - 257:45
PAC-Bayes bound for SVM (2/2) - 357:56
PAC-Bayes bound for SVM (2/2) - 459:28
PAC-Bayes bound for SVM (2/2) - 559:58
PAC-Bayes bound for SVM (2/2) - 601:00:10
PAC-Bayes bound for SVM (2/2) - 701:01:41
PAC-Bayes bound for SVM (2/2) - 801:02:10
PAC-Bayes bound for SVM (2/2) - 901:02:11
PAC-Bayes bound for SVM (2/2) - 1001:02:12
PAC-Bayes bound for SVM (2/2) - 1101:02:55
PAC-Bayes bound for SVM (2/2) - 1201:02:59
PAC-Bayes bound for SVM (2/2) - 1301:03:01
Learning the prior (1/3) - 101:03:04
Learning the prior (1/3) - 201:03:23
Learning the prior (1/3) - 301:03:44
Learning the prior (1/3) - 401:04:28
Learning the prior (1/3) - 501:04:33
New prior for the SVM (3/3) - 101:04:34
New prior for the SVM (3/3) - 201:04:43
New prior for the SVM (3/3) - 301:04:53
New prior for the SVM (3/3) - 401:05:10
New bound for the SVM (2/3) - 101:05:32
New bound for the SVM (2/3) - 201:05:39
New bound for the SVM (2/3) - 301:05:41
New bound for the SVM (2/3) - 401:05:43
New bound for the SVM (2/3) - 501:05:59
New bound for the SVM (2/3) - 601:06:08
New bound for the SVM (2/3) - 701:06:13
New bound for the SVM (2/3) - 801:07:01
Model selection with the new bound: Setup - 101:07:03
Model selection with the new bound: Setup - 201:07:29
Model selection with the new bound: Setup - 301:07:30
Model selection with the new bound: Setup - 401:07:50
Model selection with the new bound: Setup - 501:07:59
Model selection with the new bound: Results01:08:22
Tightness of the new bound01:09:09
Prior-SVM - 101:09:50
Prior-SVM - 201:10:24
Prior-SVM - 301:10:28
Prior-SVM - 401:10:50
Bound for p-SVM - 101:10:54
Bound for p-SVM - 201:11:02
Bound for p-SVM - 301:11:05
Bound for p-SVM - 401:11:11
Prior-SVM - 501:11:16
Prior-SVM - 601:11:52
Prior-SVM - 701:11:53
Prior-SVM - 801:11:59
Bound for prior-SVM - 101:12:22
Bound for prior-SVM - 201:12:23
Bound for prior-SVM - 301:12:24
Model selection with p-SVM01:13:50
Tightness of the bound with p-SVM01:14:23
Concluding remarks - 101:16:52
Concluding remarks - 201:17:21
Concluding remarks - 301:17:37
Concluding remarks - 401:17:41
Concluding remarks - 501:18:05
Concluding remarks - 601:18:11
Concluding remarks - 701:18:20