Blackwell Approachability and No-Regret Learning are Equivalent  thumbnail
Pause
Mute
Subtitles
Playback speed
0.25
0.5
0.75
1
1.25
1.5
1.75
2
Full screen

Blackwell Approachability and No-Regret Learning are Equivalent

Published on Aug 02, 20115890 Views

We consider the celebrated Blackwell Approachability Theorem for two-player games with vector payoffs. Blackwell himself previously showed that the theorem implies the existence of a “no regret” algor

Related categories

Chapter list

Blackwell Approachability <==> No-Regret Online Learning00:00
David Blackwell00:24
Review: Online Linear Optimization01:10
Online Linear Optimization01:26
Problem Solved...02:23
Blackwell Approachability03:05
Warmup: Von Neumann Minimax Theorem03:19
Von Neumann Ver. 2.004:13
Blackwell’s question05:39
NOOOOOOOO!!!!!06:52
Definitions07:50
Old Example09:25
Uses of Approachability (a very small sample)10:39
Our Main Result - 111:11
Our Main Result - 211:46
OLO Regret Minimization - 111:59
OLO Regret Minimization - 212:48
NoRegret => Blackwell13:30
Trick: Cone Duality13:46
Key Lemma15:15
Applications + Conclusions16:05
Calibrated Forecasting16:52
First Efficient Alg for Binary Calibration17:19
Thanks18:24