K. P. Naveen

Subhojyoti Mukherjee, K. P. Naveen, Nandan Sudarsanam et al.

We propose a novel variant of the UCB algorithm (referred to as Efficient-UCB-Variance (EUCBV)) for minimizing cumulative regret in the stochastic multi-armed bandit (MAB) setting. EUCBV incorporates the arm elimination strategy proposed in UCB-Improved \citep{auer2010ucb}, while taking into account the variance estimates to compute the arms' confidence bounds, similar to UCBV \citep{audibert2009exploration}. Through a theoretical analysis we establish that EUCBV incurs a \emph{gap-dependent} regret bound of {\scriptsize $O\left( \dfrac{Kσ^2_{\max} \log (TΔ^2 /K)}Δ\right)$} after $T$ trials, where $Δ$ is the minimal gap between optimal and sub-optimal arms; the above bound is an improvement over that of existing state-of-the-art UCB algorithms (such as UCB1, UCB-Improved, UCBV, MOSS). Further, EUCBV incurs a \emph{gap-independent} regret bound of {\scriptsize $O\left(\sqrt{KT}\right)$} which is an improvement over that of UCB1, UCBV and UCB-Improved, while being comparable with that of MOSS and OCUCB. Through an extensive numerical study we show that EUCBV significantly outperforms the popular UCB variants (like MOSS, OCUCB, etc.) as well as Thompson sampling and Bayes-UCB algorithms.

K. P. Naveen, Anurag Kumar, Eitan Altman

We consider a setting in which two nodes (referred to as forwarders) compete to choose a relay node from a set of relays, as they ephemerally become available (e.g., wake up from a sleep state). Each relay, when it arrives, offers a (possibly different) "reward" to each forwarder. Each forwarder's objective is to minimize a combination of the delay incurred in choosing a relay and the reward offered by the chosen relay. As an example, we develop the reward structure for the specific problem of geographical forwarding over a network of sleep-wake cycling relays. We study two variants of the generic relay selection problem, namely, the completely observable (CO) case where, when a relay arrives, both forwarders get to observe both rewards, and the partially observable (PO) case where each forwarder can only observe its own reward. Formulating the problem as a two person stochastic game, we characterize solution in terms of Nash Equilibrium Policy Pairs (NEPPs). For the CO case we provide a general structure of the NEPPs. For the PO case we prove that there exists an NEPP within the class of threshold policy pairs. We then consider the particular application of geographical forwarding of packets in a shared network of sleep-wake cycling wireless relays. For this problem, for a particular reward structure, using realistic parameter values corresponding to TelosB wireless mote, we numerically compare the performance (in terms of cost to both forwarders) of the various NEPPs and draw the following key insight: even for moderate separation between the two forwarders, the performance of the various NEPPs is close to the performance of a simple strategy where each forwarder behaves as if the other forwarder is not present. We also conduct simulation experiments to study the end-to-end performance of the simple forwarding policy.