Convexity and Information ElicitationHow can we score a weather forecaster so she gives the most accurate predictions?
Glenn Brier's answer to this question in 1950 founded the field of elicitation: incentivizing strategic agents to reveal information.
The answers have deep connections to convexity, a fundamental feature that comes up all the time in machine learning and algorithms.
This series is based on my tutorial at EC '16 with Rafael Frongillo. You can see the slides from that tutorial at the above link.
- 2016-09-20: Convexity
- 2016-09-22: Proper Scoring Rules
- 2017-04-12: Eliciting Properties
- 2017-04-22: Eliciting Finite Properties
- 2017-10-03: Prediction Markets
- 2018-03-16: Eliciting Continuous Scalars
- 2018-08-02: Eliciting Means
- 2018-08-08: Prediction Markets II
Value of Information and DecisionmakingHow to formalize the value of a piece of information for purposes of decision or prediction?
- 2016-09-24: Generalized Entropies and the Value of Information
- 2016-10-02: Risk Aversion and Max Entropy
- 2016-10-07: Divergences and Value of Information
- 2017-09-28: Risk Aversion and Decisionmaking
- 2018-07-20: Weitzman's Pandora's Box Problem
MathMath is cool, fun, and important.
This series covers useful or interesting posts that are purely about mathematics.
ProbabilityProbability is so much fun that it gets its own section.
- 2017-01-06: k-way Collisions of Balls in Bins
- 2017-10-07: Intro to Measure Concentration
- 2017-12-18: Subgaussian Variables and Concentration
- 2018-03-17: Tight Bounds for Gaussian Tails and Hazard Rates
- 2018-08-25: Prophet Inequalities
- 2018-09-29: Measure Concentration II
- 2019-02-03: Sub-Gamma Variables and Concentration