Thompson Sampling for Contextual Bandits with Linear Payoffs · HackerPapers discussions

Thompson Sampling for Contextual Bandits with Linear Payoffs — mentioned by feral

Mon, 20 Jan 2014 22:38:46 GMT

>The regret bounds on contextual bandits are O(d sqrt(T)), where d is the dimensionality of the user profile you're using and T is time (see e.g. http://arxiv.org/abs/1209.3352). In contrast, the standard stochastic bandit has regret bounds of O(log(T)), which is much tighter. The summary is you need a lot more data to explore the bigger range of possibilities you've created. So I'm not sure contextual bandits are viable for smaller sites, but Wikipedia definitely has the traffic to make them work. Its definitely viable for smaller sites. Your argument there is about the bounds in a…

Paper and discussion history. A mention is not an endorsement.

Thompson Sampling for Contextual Bandits with Linear Payoffs — mentioned by noelwelsh

Mon, 20 Jan 2014 21:01:33 GMT

I'm a big proponent of bandit algorithms (I'm a founder of Myna http://mynaweb.com/, something of a competitor to Synference). I think there is a bit of nuance here that is missing in the discussion. Should Wikipedia use a contextual bandit system? Assuming the cost is equivalent, I say yes, for the following reasons: - they have a shedload of traffic, and then some; and - they have a limited time in which the campaign is running. Let's address these in turn. The regret bounds on contextual bandits are O(d sqrt(T)), where d is the dimensionality of the user profile you're using and T is time…

Paper and discussion history. A mention is not an endorsement.