Passive Learning with Target Risk thumbnail
Pause
Mute
Subtitles
Playback speed
0.25
0.5
0.75
1
1.25
1.5
1.75
2
Full screen

Passive Learning with Target Risk

Published on Aug 09, 20132807 Views

In this paper we consider learning in passive setting but with a slight modification. We assume that the target expected loss, also referred to as target risk, is provided in advance for learner as pr

Related categories

Chapter list

Passive Learning with Target Loss00:00
Statistical Learning Theory00:12
Empirical Risk Minimization (ERM)01:00
Property Testing of Learning01:47
Outline02:24
Lower Bounds02:39
Fast and Optimistic Sample Complexities02:55
Main Result on Sample Complexity03:43
Assumptions04:02
Convex Learnability and The Curse of Stochastic Oracle04:22
Learning without Uniform Convergence04:33
Stochastic Optimization for Risk Minimization05:09
Intuition: The Curse of Stochastic Oracle05:43
SGD with Target Risk - 106:59
Three Pillars07:04
Clipping the Stochastic Gradients07:37
Shrinking the Hypothesis Space H07:59
SGD with Target Risk - 208:07
Convergence Rate09:05
Sample Complexity09:19
Proof Sketch I09:36
Proof Sketch II10:09
Conclusions and Open Problems10:59
Thank you13:05