@avibryant have you seen this: http://sifter.cc/ something @adamlaiacano made to use bandits to do testing.
@posco I hadn't, thanks. @adamlaiacano is it contextual or does it assume visitors are uniform?
-
-
@avibryant For adaptive testing with targeting you can use@conductrics. Let me know and I can set you up.http://conductrics.com/intelligent-agents-ab-testing-user-targeting-and-predictive-analytics/ … -
@mgershoff@conductrics frank feedback: the post you linked is probably great for marketers but the "agent" stuff leaves this engineer cold. -
@avibryant ha, no worries. How about http://conductrics.com/data-science-resources-2 … ? Anyway you can do contextual bandits and multi-step optimization via API. -
@mgershoff much better, thank you :) -
@avibryant Actually, maybe this is what you are looking for http://console.conductrics.com/docs/demo It has the API docs and an overview of the service.
End of conversation
New conversation -
-
-
@avibryant@posco assumes users are uniform, but you could weigh the reward in the client side. -
@adamlaiacano nice API, have you considered adding in "stochastic MAB" methods? UCB, Thompson sampling, etc.? -
@jamneuf I have UCB/ucb2 algos in the can but haven't deployed yet. Aren't they less stochastic? Arm used is deterministic given the state. -
@adamlaiacano yep, "stochastic" here refers to the domain assumptions, I'de advocate TS over UCB, some reasons here: http://simplemlhacks.blogspot.ca/2013/04/reddits-best-comment-scoring-algorithm.html … -
@jamneuf I was just skimming that on my phone. Looking forward to a full read. Have you read@johnmyleswhite's bandits book? Solid MAB intro -
@adamlaiacano@johnmyleswhite nope, looks pretty cool though, good to know -
@jamneuf@adamlaiacano Nice post. Hadn't seen the finite sample proof from Kaufmann. I can never tell how TS depends on choice of prior. -
@johnmyleswhite@adamlaiacano no one can, but when are people ever okay with priors, JLangford has a good rant about this, can't find now
End of conversation
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.