On the Optimal Sample Complexity for Best Arm Identification
This work addresses the sample efficiency challenge in multi-armed bandits for researchers and practitioners, though it is incremental as it builds on existing lower bounds and algorithms.
The paper tackles the best arm identification problem in stochastic bandits by developing a nontrivial algorithm that improves prior upper bounds and establishing a new lower bound for the two-arm case, which extends beyond classic results.
We study the best arm identification (BEST-1-ARM) problem, which is defined as follows. We are given $n$ stochastic bandit arms. The $i$th arm has a reward distribution $D_i$ with an unknown mean $μ_{i}$. Upon each play of the $i$th arm, we can get a reward, sampled i.i.d. from $D_i$. We would like to identify the arm with the largest mean with probability at least $1-δ$, using as few samples as possible. We provide a nontrivial algorithm for BEST-1-ARM, which improves upon several prior upper bounds on the same problem. We also study an important special case where there are only two arms, which we call the sign problem. We provide a new lower bound of sign, simplifying and significantly extending a classical result by Farrell in 1964, with a completely new proof. Using the new lower bound for sign, we obtain the first lower bound for BEST-1-ARM that goes beyond the classic Mannor-Tsitsiklis lower bound, by an interesting reduction from Sign to BEST-1-ARM. We propose an interesting conjecture concerning the optimal sample complexity of BEST-1-ARM from the perspective of instance-wise optimality.