Andy Zheng

PhD Student in Operations Research at MIT. H-index=0 but Emacs config over 9000 (lines)

[Andy Zheng @atzheng]

New on arxiv: some fun work we did training an RL agent to play Liar’s Poker!

[Richard Dewey @richdewey]

1/ Paper alert: Introducing Solly, the first AI to achieve elite human level performance at Liar’s Poker, the classic Wall Street bluffing game. With @Ciamac, @atzheng and Janos Botyanszki.

[Andy Zheng @atzheng]

@clcrozier @jeremyphoward Great illustration! You might enjoy our simple analysis of this problem https://t.co/Kylxpo2pug

[Andy Zheng @atzheng]

@sam_power_825 @_jcken The best shrinkage estimator (in MSE) is biased for any x, no? I.e., argmin_{α ≥ 0} { E |α y - x|² } < 1. Maybe the fun part is then showing that James-Stein actually allows you to choose a shrinkage α based on y, which will be better than 1

[Andy Zheng @atzheng]

Where has this been my whole life?? Magically render arxiv papers as HTML in your browser: https://t.co/FNUVsh5QO4

[Andy Zheng @atzheng]

@kareem_carr JuMP is the best interface I've seen to solvers for MIPs, LPs, SOCPs, and more https://t.co/ye7GYxpJSj

[Andy Zheng @atzheng]

@johncarlosbaez Some great geometric intuition in high dimensions: James-Stein approximates the projection of the mean onto the sample mean. https://t.co/a7mF1VOPIs

Also, little-known but obvious in hindsight: you can derive the James-Stein estimator by minimizing SURE! https://t.co/92Rnxk3G6S