As an alternative to stochastic approximation, you can also use an active learning framework. See e.g. Section III of http://www.cs.ubc.ca/~nando/papers/aplrss.pdf (apologies for the self-promotion) and references [13], [15], [29], [30] and [32] in this paper. Good luck. Nando.
[EMAIL PROTECTED] wrote: > > You might want to take a look at this book: > > Adaptive Algorithms and Stochastic Approximations. > Benveniste, Metivier and Prioutet. 1990. > > -- Chunnan > > > > Quoting Ronen Brafman <[EMAIL PROTECTED]>: > >> I want to find the maximum of f(x1,.,x-n). n is typically not too large. >> >> I can only sample the value of f, and the samples are noisy. >> >> >> >> I wonder if anyone can provide pointers to results that study how >> this can >> be done under various assumptions on f. >> >> >> >> Thanks, >> >> >> >> Ronen >> >> > > > > ---------------------------------------------------------------- > This message was sent using IMP, the Internet Messaging Program. > >------------------------------------------------------------------------ > >_______________________________________________ >uai mailing list >uai@ENGR.ORST.EDU >https://secure.engr.oregonstate.edu/mailman/listinfo/uai > > _______________________________________________ uai mailing list uai@ENGR.ORST.EDU https://secure.engr.oregonstate.edu/mailman/listinfo/uai
