An Asymptotically Optimal Policy for Uniform Bandits of Unknown Support
This provides an asymptotically optimal solution for bandit problems with uniform distributions, which is incremental as it builds on existing lower bounds and policies.
The paper tackles the problem of sequential sampling from multiple uniform distributions with unknown support to maximize cumulative outcomes, presenting an inflated sample mean policy that achieves asymptotic optimal regret matching the known lower bound.
Consider the problem of a controller sampling sequentially from a finite number of $N \geq 2$ populations, specified by random variables $X^i_k$, $ i = 1,\ldots , N,$ and $k = 1, 2, \ldots$; where $X^i_k$ denotes the outcome from population $i$ the $k^{th}$ time it is sampled. It is assumed that for each fixed $i$, $\{ X^i_k \}_{k \geq 1}$ is a sequence of i.i.d. uniform random variables over some interval $[a_i, b_i]$, with the support (i.e., $a_i, b_i$) unknown to the controller. The objective is to have a policy $π$ for deciding, based on available data, from which of the $N$ populations to sample from at any time $n=1,2,\ldots$ so as to maximize the expected sum of outcomes of $n$ samples or equivalently to minimize the regret due to lack on information of the parameters $\{ a_i \}$ and $\{ b_i \}$. In this paper, we present a simple inflated sample mean (ISM) type policy that is asymptotically optimal in the sense of its regret achieving the asymptotic lower bound of Burnetas and Katehakis (1996). Additionally, finite horizon regret bounds are given.