A Collaborative Mechanism for Crowdsourcing Prediction Problems thumbnail
Pause
Mute
Subtitles
Playback speed
0.25
0.5
0.75
1
1.25
1.5
1.75
2
Full screen

A Collaborative Mechanism for Crowdsourcing Prediction Problems

Published on Jan 25, 20127840 Views

Machine Learning competitions such as the Netflix Prize have proven reasonably successful as a method of “crowdsourcing” prediction tasks. But these competitions have a number of weaknesses, particula

Related categories

Chapter list

A collaborative mechanism for crowdsourcing prediction problems00:00
De Finetti: "Probability does not exist"00:25
Intrade prediction market01:47
Markets make good predictors02:21
What to do when the market is thin?02:48
Contract pricing via automated market maker03:22
Logarithmic market scoring rule04:37
Interlude: prize-driven competitions06:00
Prizes everywhere ... - 106:24
Prizes everywhere ... - 206:53
First place team07:35
2nd place finisher: "Grand prize team"08:19
Problems with "Netflix-Style" competitions08:47
Our proposal: crowdsourced learning mechanism09:42
Ingredient I: loss function10:04
Ingredient II: learning mechanism10:47
Ingredient II: correct incentives12:03
Result: equivalent to a share-trading market?12:41
Conclusions13:29
Thanks!14:38