Black-box Acceleration of Las Vegas Algorithms and Algorithmic Reverse Jensen's Inequalities
arXiv:2304.11017
Abstract
Let be a Las Vegas algorithm, i.e. an algorithm whose running time is a random variable drawn according to a certain probability distribution . In 1993, Luby, Sinclair and Zuckerman [LSZ93] proved that a simple universal restart strategy can, for any probability distribution , provide an algorithm executing and whose expected running time is , where is the minimum expected running time achievable with full prior knowledge of the probability distribution , and is the -quantile of . Moreover, the authors showed that the logarithmic term could not be removed for universal restart strategies and was, in a certain sense, optimal. In this work, we show that, quite surprisingly, the logarithmic term can be replaced by a smaller quantity, thus reducing the expected running time in practical settings of interest. More precisely, we propose a novel restart strategy that executes and whose expected running time is where . This quantity is, up to a multiplicative factor, better than: 1) the universal restart strategy of [LSZ93], 2) any -quantile of for , 3) the original algorithm, and 4) any quantity of the form for a large class of concave functions . The latter extends the recent restart strategy of [Zam22] achieving , and can be thought of as algorithmic reverse Jensen's inequalities. Finally, we show that the behavior of at infinity controls the existence of reverse Jensen's inequalities by providing a necessary and a sufficient condition for these inequalities to hold.
arXiv comment13 pages, 0 figures