The 26th International Conference on Machine Learning (ICML 2009)

Drifting Games, Boosting and Online Learning

author:Yoav Freund, Department of Computer Science and Engineering, University of San Diego
published: Aug. 26, 2009, recorded: June 2009, views: 331

Turn off the lights

You might be experiencing some problems with Your Video player.

Slides

Slides
0:00	Linear separation, drifting games & Boosting
0:35	Adaboost is sensitive to label noise (1)
1:52	Adaboost is sensitive to label noise (2)
2:10	Adaboost is sensitive to label noise
2:22	Robustboost - A new boosting algorithm (1)
2:58	Robustboost - A new boosting algorithm (2)
3:00	Robustboost - A new boosting algorithm (3)
3:31	Plan of talk
4:27	Label noise and convex loss functions (1)
4:35	Label noise and convex loss functions (2)
4:49	Label noise and convex loss functions (3)
5:02	Label noise and convex loss functions (4)
5:10	Label noise and convex loss functions (5)
5:25	Label noise and convex loss functions (6)
5:51	Loss functions
7:24	A hard case (1)
8:14	A hard case (2)
9:59	A hard case (3)
10:02	A hard case (4)
10:55	Plan of talk
10:56	Boost by Majority (BBM) (1)
10:57	Boost by Majority (BBM) (2)
11:11	Boost by Majority (BBM) (3)
11:23	Boost by Majority (BBM) (3)
11:59	Boost by Majority (BBM) (4)
12:04	Boost by Majority (BBM) (5)
12:10	Boost by Majority (BBM) (6)
12:28	Boost by Majority (BBM) (7)
12:32	Boost by Majority (BBM) (8)
12:47	BBM as a drifting game (1)
12:59	BBM as a drifting game (2)
13:07	BBM as a drifting game (3)
13:31	The continuous chip limit (1)
13:39	The continuous chip limit (2)
13:47	The continuous chip limit (3)
14:25	The boosting game lattice (1)
14:47	The boosting game lattice (2)
15:07	The boosting game lattice (3)
15:24	The boosting game lattice (4)
15:40	The boosting game lattice (5)
16:06	The boosting game lattice (5)
16:14	The boosting game lattice (6)
16:42	The boosting game lattice (7)
17:02	The boosting game lattice (8)
17:10	The boosting game lattice (9)ž
17:54	The boosting game lattice (10)
18:06	The boosting game lattice (11)
18:10	The boosting game lattice (12)
18:19	The boosting game lattice (13)
18:27	Weak Learner’s min/max strategy (1)
18:47	Weak Learner’s min/max strategy (2)
18:59	Weak Learner’s min/max strategy (3)
20:23	Potential (1)
20:51	Potential (2)
21:35	Potential (3)
21:51	Potential (4)
22:07	Potential (5)
22:39	Potential (7)
22:59	Potential (8)
23:07	Potential (9)
23:19	Potential (10)
23:38	Definitions (2)
23:54	Definitions (3)
24:08	Definitions (4)
24:38	Definitions (5)
24:54	Proof of Theorem (2)
25:18	Proof of Theorem (3)
25:26	Proof of Theorem (4)
25:42	Proof of Theorem (5)
25:51	Proof of Theorem (6)
25:55	Proof of Theorem (7)
26:03	Proof of Theorem (8)
26:15	Theorem about BBM (1)
26:19	Theorem about BBM (2)
26:31	Theorem about BBM (3)
26:51	t=1
27:15	t=11
27:19	t=51
27:35	t=91
27:39	t=101
27:55	BBM/Logitboost/Adaboost
28:55	High level summary (2)
29:03	High level summary (3)
29:09	High level summary (4)
29:30	Plan of talk
29:35	Why is BBM not practical? (1)
29:51	Why is BBM not practical? (2)
29:59	Why is BBM not practical? (3)
30:55	Why is BBM not practical? (5)
31:19	Why is BBM not practical? (6)
31:42	Why is BBM not practical? (7)
31:45	Why is BBM not practical? (8)
31:56	Letting time step decrease to zero (1)
31:59	Letting time step decrease to zero. (2)
32:06	Letting time step decrease to zero. (3)
32:49	Letting time step decrease to zero. (4)
33:22	Letting time step decrease to zero. (5)
34:08	Letting time step decrease to zero. (6)
34:47	Letting time step decrease to zero. (7)
35:02	The game lattice (1)
35:12	The game lattice (2)
35:28	The game lattice (3)
35:40	The game lattice (4)
35:44	The game lattice (5)
35:56	The game lattice (7)
36:04	The game lattice (8)
36:44	The game lattice (9)
36:52	The game lattice (10)
36:56	The game lattice (11)
36:59	The game lattice (13)
37:15	The game lattice (14)
37:21	The game lattice (15)
37:23	The game lattice (16)
37:38	The game lattice (17)
37:44	The game lattice (18)
38:25	Potentials in continuous time (1)
38:27	Potentials in continuous time (2)
38:37	Potentials in continuous time (3)
38:57	Example: From BBM to Brownboost (1)
39:01	Example: From BBM to Brownboost (3)
39:13	Example: From BBM to Brownboost (7)
39:29	Example: From BBM to Brownboost (8)
39:33	Example: From BBM to Brownboost (10)
40:01	Plan of talk
40:17	Robustboost (1)
40:21	Robustboost (2)
41:01	Robustboost (3)
41:28	Robustboost (4)
41:47	Robustboost (5)
41:57	Plan of talk
41:59	Experimental Results on Long/Servedio synthetic example
42:17	LogitBoost on Long/Servedio
42:21	Robustboost on Long/Servedio
43:57	JBoost V2.0
44:13	Experimental Results on real-world data
44:25	Logitboost 0% Noise (1)
44:42	Logitboost 0% Noise (2)
44:46	Logitboost 0% Noise (3)
44:49	Logitboost 0% Noise (4)
44:58	Logitboost 0% Noise (5)
45:10	Logitboost 20% Noise (1)
45:12	Logitboost 20% Noise (2)
45:15	Logitboost 20% Noise (3)
45:16	Logitboost 20% Noise (4)
45:18	Logitboost 20% Noise (5)
45:20	Logitboost 20% Noise (6)
45:46	Robustboost 20% Noise (2)
45:48	Robustboost 20% Noise (3)
45:50	Robustboost 20% Noise (4)
46:02	Robustboost 20% Noise (5)
46:14	Robustboost 20% Noise (6)
46:17	Robustboost 20% Noise (7)
46:19	Robustboost 20% Noise (8)
46:20	Robustboost 20% Noise (9)
46:23	Robustboost 20% Noise (10)
46:56	Plan of talk
46:57	For more details see my publications page (1)
47:12	For more details see my publications page (2)
47:31	For more details see my publications page (3)
47:35	For more details see my publications page (4)
47:41	For more details see my publications page (5)
47:50	Summary (1)
47:51	Summary (2)
47:57	Summary (3)
48:01	Summary (4)
48:05	Summary (5)
48:38	Summary (6)
48:51	- Questions

Report a problem or upload files

If you have found a problem with this lecture or would like to send us extra material, articles, exercises, etc., please use our ticket system to describe your request and upload the data.
Enter your e-mail into the 'Cc' field, and we will keep you updated with your request's status.

Lecture popularity: You need to login to cast your vote.