en-es
en
0.25
0.5
0.75
1.25
1.5
1.75
2
The price of bandit information in multiclass online classification
Published on 2013-08-092974 Views
We consider two scenarios of multiclass online learning of a hypothesis class H⊆YX. In the full information scenario, the learner is exposed to instances together with their labels. In the bandit scen
Related categories
Presentation
The price of bandit information in multiclass online classification00:00
Online multiclass classification: Full info vs Bandits00:13
Outline - 101:30
Hypothesis classes based online learning01:43
Setting - Mistake bounds02:20
A toy example03:05
Main Result04:30
Previous Results04:49
Outline - 206:23
The Littlestone dimension (Littlestone, 89, Daniely et al, 11)06:26
The Littlestone dimension (Littlestone, 89) - 107:44
The Littlestone dimension (Littlestone, 89) - 207:59
The Littlestone dimension (Littlestone, 89) - 308:16
The Littlestone dimension (Littlestone, 89) - 408:33
The Littlestone dimension (Littlestone, 89) - 508:45
The Littlestone dimension (Littlestone, 89) - 608:57
The Littlestone dimension (Littlestone, 89) - 709:10
The Littlestone dimension (Littlestone, 89) - 809:27
The Littlestone dimension (Littlestone, 89) - 909:31
The Littlestone dimension (Littlestone, 89) - 1009:48
The Littlestone dimension (Littlestone, 89) - 1109:56
Proof of the main theorem: a preface - 110:11
Proof of the main theorem: a preface - 211:00
Proof of the main theorem: a preface - 311:17
Proof of the main theorem: a preface - 411:18
Proof of the main theorem: a preface - 511:19
Proof of the main theorem: a preface - 611:24
Proof of the main theorem: a preface - 711:35
Proof of the main theorem: the splitting mechanism12:08
Proof of the main theorem: the algorithm12:54
Proof of the main theorem: analysis - 113:20
Proof of the main theorem: analysis - 213:56
Conclusion and open questions14:26