Uncertainty in Artificial Intelligence
First Name   Last Name   Password   Forgot Password   Log in!
    Proceedings         Authors   Author's Info   Article details         Search    
Optimal amortized regret in every interval
Rina Panigrahy, Preyas Popat
Consider the classical problem of predicting the next bit in a sequence of bits. A standard performance measure is regret (loss in payoff) with respect to a set of experts. For exam- ple if we measure performance with respect to two constant experts one that always predicts 0‚??s and another that always predicts‚?? it is 1‚??s well known that one can get regret O( T ) with respect to the best expert by using, say, the weighted majority algorithm [LW89]. But this algorithm does not provide performance guaran- tee in any interval. There are other algorithms (see [BM07, FSSW97, Vov99]) that ensure regret ‚?? O( x log T ) in any interval of length x. In this paper we show a randomized algorithm that in an ‚?? amortized sense gets a regret of O( x) for any interval when the sequence is partitioned into in- tervals arbitrarily. We empirically estimated the constant in the O() for T upto 2000 and found it to be small ‚?? around 2.1. We also experimentally evaluate the efficacy of this algorithm in predict- ing high frequency stock data.
Pages: 663-671
PS Link:
PDF Link: /papers/14/p663-panigrahy.pdf
AUTHOR = "Rina Panigrahy and Preyas Popat",
TITLE = "Optimal amortized regret in every interval",
BOOKTITLE = "Proceedings of the Thirtieth Conference Annual Conference on Uncertainty in Artificial Intelligence (UAI-14)",
ADDRESS = "Corvallis, Oregon",
YEAR = "2014",
PAGES = "663--671"

hosted by DSL   •   site info   •   help