Nonstochastic bandits: Countable decision set, unbounded costs and reactive environments.

Jan Poland Nonstochastic bandits: Countable decision set, unbounded costs and reactive environments. 77-93 2008 397 Theor. Comput. Sci. 1-3 http://dx.doi.org/10.1016/j.tcs.2008.02.024 db/journals/tcs/tcs397.html#Poland08