Delay as Payoff in MAB
This work addresses a novel problem in sequential decision-making for applications like networking and web content, offering theoretical improvements but is incremental in the broader MAB literature.
The paper tackles a variant of the stochastic Multi-armed Bandit problem where payoff is delayed and proportional to delay, modeling scenarios like network routing or user engagement, and provides tight upper and lower regret bounds, such as optimal regret scaling as ∑(log T/Δ_i) + d* for cost settings and ∑(log T/Δ_i) + d̄ for reward settings, improving over general delay-dependent bounds.
In this paper, we investigate a variant of the classical stochastic Multi-armed Bandit (MAB) problem, where the payoff received by an agent (either cost or reward) is both delayed, and directly corresponds to the magnitude of the delay. This setting models faithfully many real world scenarios such as the time it takes for a data packet to traverse a network given a choice of route (where delay serves as the agent's cost); or a user's time spent on a web page given a choice of content (where delay serves as the agent's reward). Our main contributions are tight upper and lower bounds for both the cost and reward settings. For the case that delays serve as costs, which we are the first to consider, we prove optimal regret that scales as $\sum_{i:Δ_i > 0}\frac{\log T}{Δ_i} + d^*$, where $T$ is the maximal number of steps, $Δ_i$ are the sub-optimality gaps and $d^*$ is the minimal expected delay amongst arms. For the case that delays serves as rewards, we show optimal regret of $\sum_{i:Δ_i > 0}\frac{\log T}{Δ_i} + \bar{d}$, where $\bar d$ is the second maximal expected delay. These improve over the regret in the general delay-dependent payoff setting, which scales as $\sum_{i:Δ_i > 0}\frac{\log T}{Δ_i} + D$, where $D$ is the maximum possible delay. Our regret bounds highlight the difference between the cost and reward scenarios, showing that the improvement in the cost scenario is more significant than for the reward. Finally, we accompany our theoretical results with an empirical evaluation.