The Multi-Armed Bandit Problem

http://camdp.com/blogs/multi-armed-bandits

by zenburnyourface • 13 years ago

7 9 13 years ago

camdp.com

spitfire2• 13 years ago

Good article.

I'd like to know what he used to output those graphs.

zenburnyourface• 13 years ago

It uses Python's Matplotlib with a custom styling, available here: https://github.com/CamDavidsonPilon/Probabilistic-Programmin...

spitfire2• 13 years ago

Thanks! I quite like the output. Are you the author?

If soI took a quick look at your book and it looks excellent.

https://github.com/CamDavidsonPilon/Probabilistic-Programmin...

bun-neh• 13 years ago

One thing to expand upon: This is just the classic formulation of the Multi-Armed Bandit Problem. There are countless variations to the formula which have slightly different underlying assumptions.

mcpherrinm• 13 years ago

Eugh, this website has custom behavior for the left-right arrow keys, which breaks the back keyboard shortcut in Firefox (alt+left arrow key). Non-standard navigation on a page that IMHO doesn't need it.

triplesec• 13 years ago

spacebar scrolling also broken in ff. grrrrr

zenburnmyface2• 13 years ago

haha ironically left-arrow right-arrow replace space-bar scroll.

mailshanx• 13 years ago

Just had a look at your book, and it looks amazing! Hope you will complete it:D

On a separate note, im surprised that in the many discussion of the MAB problem on HN, the Gittins index was never mentioned. Under certain (mild) assumptions on reward distributions, the GI gives a provably optimal solution to the MAB problem...

The Multi-Armed Bandit Problem

Related Stories